top of page
Publications
Healthy City Lab / Publications
Below is a list of publications from the Healthy City Lab. For the most current and complete record, see Dr. Bayat's Google Scholar profile.
Bold indicates Dr. Bayat. Underlined indicates trainees in the lab.
A framework for naturalistic driving assessment in people living with dementia
Hettiarachchige RO, Naglie G, Rapoport MJ, Alizadeh S, Bayat S
Gerontology
Naturalistic Driving Studies (NDSs) offer a promising approach for assessing driving safety in older adults living with mild dementia. Traditional tools, such as on-road tests and driving simulators, often fall short in capturing everyday driving behaviour, as they are conducted in structured, time-limited, and sometimes stressful settings. In contrast, NDS can monitor real-world driving over extended periods and in natural contexts. This commentary reviews the current landscape of NDS involving individuals with mild dementia, outlining key limitations in study design, including heterogeneous sensor configurations, inconsistent driving metrics, and a lack of standard outcome definitions. To address these issues, we propose a next-generation framework for NDS design that emphasizes core sensor technologies such as inertial measurement unit, global positioning system, and cameras, and highlights the importance of integrated, resource-efficient systems. We offer recommendations to enhance the validity, comparability, and clinical utility of naturalistic driving methods in clinical research. Although the framework focuses on mild dementia, its principles may also be applied to other populations experiencing cognitive impairment.
Sleep Timing and Disruption as Candidate Digital Biomarkers Along a Cognitive Risk Continuum: A Feasibility Study Using Actigraphy
Chatterjee R, Aboudeif S, Denroche B, Gill B, Ghayur Sadigh A, Barber PA, Bayat S
IEEE EMBC 2026, Toronto, Canada
Actigraphy is a non-invasive, rest-activity cycles monitoring method using wrist‑worn accelerometers. GGIR is an open‑source R package that processes raw accelerometry to derive sleep-wake metrics. This pilot study assessed the feasibility of actigraphy‑derived sleep measures for cognitive risk assessment across healthy controls (HC), transient ischemic attack (TIA), and Alzheimer’s disease (AD), and evaluated whether observed patterns aligned with known sleep-cognition relationships. Twelve older adults (71.3 ± 7.5 years; 3 females; HC = 6, TIA = 3, AD = 3) completed ~2 weeks of wrist actigraphy across two at‑home monitoring periods. GGIR was used to extract nightly wake after sleep onset (WASO), number of awakenings, sleep duration, mid‑sleep timing and variability in sleep duration and mid‑sleep. Valid nights required >80% usable data. Spearman correlations examined associations with Montreal Cognitive Assessment (MoCA), and Kruskal-Wallis tests assessed group differences in variability. Participants contributed 15.8 valid nights on average. Greater sleep disruption related to lower cognition: WASO (ρ ≈−0.69) and awakenings (ρ≈−0.68) showed strong negative trends with MoCA. Sleep mid‑point showed a weaker negative trend (ρ≈−0.31), and duration a weak positive trend (ρ≈0.29). Variability in timing and duration increased along the HC→ TIA → AD continuum. Automated sleep‑period detection underperformed during prolonged nocturnal wake. Actigraphy processed with GGIR provides interpretable sleep features and reveals graded patterns across cognitive risk groups, supporting its
GeoFuse: A Scalable Multimodal Greenspace Quantification and Fusion Toolbox
Ghayur Sadigh A, Stefanakis E, Bayat S
IEEE EMBC 2026, Toronto, Canada
Quantifying urban greenspace is foundational to environmental health research, yet widely used satellite metrics such as NDVI capture canopy biomass from a top-down view and often fail to reflect the vertical greenery experienced by pedestrians. Conversely, street-level measures like the Green View Index (GVI) provide a human-centric perspective but are challenging to compute at city scale due to data acquisition costs. To bridge this gap, we present GeoFuse, an open-source toolbox that automates multimodal greenspace profiling. The system (i) retrieves Sentinel-2 imagery via Google Earth Engine to quantify NDVI, (ii) extracts GVI from Google Street View using semantic segmentation, and (iii) fuses modalities into an outcome-specific "Composite Greenery Index" (CGI) by optimizing weights against health targets. Engineered for cityscale processing, GeoFuse utilizes a distributed backend based on Message Passing Interface (MPI) for High-Performance Computing. In a Calgary case study, the toolbox identified distinct drivers; respiratory health was associated with total biomass (NDVI weight ≈80%), whereas diabetes prevalence was driven by visible street vegetation (Vegetation weight ≈55%). The optimized CGI demonstrated stronger associations than standalone metrics (e.g., Diabetes: r = -0.595 vs -0.589). By automating the discovery of these disease-specific exposure profiles, GeoFuse offers a reproducible framework for precision public health research.
Predicting Depression Using Physical Activity, Sleep, and Demographic Factors in NHANES: A Fairness-Centred Machine Learning Approach
Wallich M, Bento M, Bayat S
IEEE EMBC 2026, Toronto, Canada
Depression is a prevalent mental health condition influenced by behavioural, demographic, and social factors, yet it often remains underdiagnosed due to reliance on subjective assessments. This study evaluated whether objectively measured behavioural features from accelerometers can predict depression while maintaining fairness across demographic subgroups. We trained and compared multiple machine learning models using sleep and physical activity features, with and without demographic variables, and assessed performance and bias using F1- score, sensitivity, specificity, and fairness metrics. Random forest models achieved the best performance, with a maximum F1-score of 83%. The addition of demographic variables improved the F1- score; however, the bias across sensitive groups increased, while behavioural features alone maintained lower subgroup disparities with lower performance (F1 = 0.77). These results suggest that the inclusion of accelerometer-derived behavioural features can support more equitable depression prediction.
The Neighbourhood Built Environment Affects Driving Behaviours of Older Adults: A Combined Geographic Information Systems and Machine Learning Method
Hafezifar R, Alizadeh S, Dickerson A, Vrkljan B, Babulal GM, Bayat S
Cities & Health
Driving space is considered as the transaction between built environment features and driving behaviour. Driving keeps people active and engaged, particularly in later life. Using Geospatial Information Systems (GIS) and machine learning, this study examined the driving space of older drivers (aged ≥65; n = 134) living in St. Louis City, St. Louis County, USA from 1 January 2019, to 31 December 2019. Driving variables, such as total distance, trip frequency, ratio of short trips long trips, were analyzed. Built environment measures included transit accessibility, land use mix, and road network characteristics. Our findings indicate that the most important features predictive of driving space of older adults were public transit density and land use diversity within residential areas. This study demonstrates the non-linear relationship between built environment factors and driving space variables. Total distance has a complex relationship with each built environment variable. The differences in short-distance and long-distance driving are linked to varied land use types, balanced transport density, and intersection density. These findings highlight the value of using in-vehicle monitoring technologies to determine how specific characteristics of the built environment can influence everyday driving behaviours in later life.
Accuracy-fairness trade-off in ML for healthcare: A quantitative evaluation of bias mitigation strategies
Dehghani F, Paiva P, Malik N, Lin J, Bayat S, Bento M
Information and Software Technology
Context: Although machine learning (ML) has significant potential to improve healthcare decision-making, embedded biases in algorithms and datasets risk exacerbating health disparities across demographic groups. To address this challenge, it is essential to rigorously evaluate bias mitigation strategies to ensure fairness and reliability across patient populations.
Objective: The aim of this research is to propose a comprehensive evaluation framework that systematically assesses a wide range of bias mitigation techniques at pre-processing, in-processing, and post-processing stages, using both single- and multi-stage intervention approaches.
Methods: This study evaluates bias mitigation strategies across three clinical prediction tasks: breast cancer diagnosis, stroke prediction, and Alzheimer’s disease detection. Our evaluation employs group- and individual-level fairness metrics, contextualized for specific sensitive attributes relevant to each dataset. Beyond fairness-accuracy trade-offs, we demonstrate how metric selection must align with clinical goals (e.g., parity metrics for equitable access, confusion-matrix metrics for diagnostics).
Results: Our results reinforce that no single classifier or mitigation strategy is universally optimal, underscoring the value of our proposed framework for evaluating fairness and accuracy throughout the bias mitigation process. According to the results, Adversarial Debiasing improved fairness by 95% in breast cancer diagnosis without compromising accuracy. Reweighing was most effective in stroke prediction, boosting fairness by 41%, and Reject Option Classification yielded nearly 50% fairness improvement in Alzheimer’s detection. Multi-stage bias mitigation did not consistently lead to better outcomes, and in many cases, fairness gains came at the expense of accuracy.
Conclusion: These findings provide practical guidance for selecting fairness-aware machine learning strategies in healthcare, aiding both model development and benchmarking across diverse clinical applications.
Developing a Route Complexity Metric for Trajectories: A Case Study on Driving Behaviours of Older Adults with Preclinical Alzheimer's Disease
Long K, Babulal G, Bayat S
IEEE Journal of Translational Engineering in Health and Medicine
Objective: To examine how early pathophysiological changes in Alzheimer’s disease (AD) affect navigational decision-making by analyzing the complexity of driving routes in older adults with and without preclinical AD. Methods: We developed a novel route complexity metric based on the number of left and right turns and the deviation from the most direct path, accounting for cognitive load during navigation. Naturalistic GPS driving data were collected for a year from 111 older adults aged 65–85, with preclinical AD status determined via cerebrospinal fluid amyloid biomarkers. A multiple linear regression model was used to assess the relationship between age, preclinical AD status, and route complexity. Results: The findings of this study indicate that preclinical AD may influence the navigational abilities of older adults. After controlling for age, participants with preclinical AD chose routes with higher baseline complexity than the control group. It further revealed that participants with preclinical AD selected routes with lower complexity as they aged—a trend not observed in healthy controls. Conclusion: Preclinical AD is associated with changes in spatial decision-making that are observable in real-world driving behaviours. The age-related decline in route complexity among those with preclinical AD may reflect compensatory strategies or progressive cognitive changes. Clinical Impact: This study presents a non-invasive, behaviour-based metric that could support early detection of cognitive decline. It may also inform the design of personalized mobility interventions and dementia-friendly mobility systems.
Rule-Based Detection of Turns and Curves in Naturalistic Driving Using GPS and Gyroscope Data
Hassanin O, Vrkljan B, Bayat S
IEEE Sensors Letters
This letter presents an interpretable, rule-based framework for detecting and classifying common driving maneuvers using naturalistic sensor data collected from older adult drivers. The method relies solely on GPS-derived heading and gyroscope-based angular velocity, avoiding traffic-sensitive variables such as speed and acceleration. A two-step approach was implemented: thresholding gyroscope signals to detect sharp maneuvers and analyzing monotonic trends in GPS heading to capture gradual maneuvers. Each detected maneuver was then classified into distinct categories—loops, 90° turns, and curves (tight/wide × smooth/sharp)—using features such as heading change, peak angular velocity, and spatial extent. Evaluation on over 500 annotated events showed a classification accuracy of 98.6%, with high performance across most maneuver types. The framework is sensor-efficient, robust to driving variability, and well-suited for real-world applications in driver behavior monitoring and safety assessment.
Estimating movement direction from body orientation using dual ultra-wideband sensor configuration
Ojghaz A, Bayat S, Sadeghpour F
IEEE Sensors Letters
Accurate short-term prediction of human movement is vital for safety-critical and context-aware applications in dynamic environments. While conventional trajectory prediction methods depend on historical motion data, they often fall short in anticipating sudden directional changes. This study investigates whether body orientation, estimated using a dual ultra-wideband (UWB) sensor configuration, can serve as a reliable predictor of near-future movement direction. A wearable device with two shoulder-mounted UWB tags was used to collect position and orientation data during controlled walking experiments. Eight participants walked freely within a controlled lab environment while data were recorded. Circular cross-correlation was applied to analyze the temporal relationship between body orientation and subsequent movement direction. Results revealed a strong and statistically significant correlation across all participants (mean correlation = 0.7688, p
Common driving behaviors in older adults with dementia: Insights from a systematic literature review
Hettiarachchige RH, Rapoport MJ, Naglie G, Vingilis E, Seeley J, Alizadeh S, Bayat S
Alzheimer's & Dementia
Dementia impairs driving skills, but the specific driving behaviors affected are not fully understood. This project reviewed the literature on driving behaviors more common among people with dementia compared to age-matched healthy controls. A search of Scopus, Medline All, and Embase databases (1994 to September 2024) identified relevant studies. Articles were included if they addressed driving behaviors among drivers with dementia during on-road tests, simulator experiments, or naturalistic driving, and included comparisons with non-dementia controls. Of 2359 citations, 26 studies were included: 3 used naturalistic driving, 14 driving simulators, and 9 used on-road tests. Drivers with dementia showed higher standard deviations of mean speeds, more traffic light tickets, greater out-of-lane drifting, and increased variability in mean headway distance compared to controls. Findings highlight distinct driving behavior patterns among drivers with dementia. However, these results should be interpreted cautiously due to methodological limitations, including small samples, lack of confounding factors, and non-validated settings. HIGHLIGHTS: Drivers with dementia exhibit distinct driving patterns that consistently set them apart from cognitively intact drivers. Compared to age-matched controls, drivers with dementia are more likely to demonstrate greater variability in mean speeds, accumulate more traffic light violations, exhibit higher instances of lane drifting, and show increased variability in mean headway distance. Driving simulators, on-road tests, and naturalistic driving methods have been used to study driving in individuals with dementia, although most evidence comes from simulator studies, which may not fully reflect real-world driving conditions.
Identifying major depressive disorder in older adults through naturalistic driving behaviors and machine learning
Chen C, Brown DC, Al Hammadi N, Bayat S, Dickerson A, Vrkljan B, Blake M, Zhu Y, Trani J-F, Lenze EJ, Carr DB, Babulal GM
Nature Digital Medicine
Depression in older adults is often underdiagnosed and has been linked to adverse outcomes, including motor vehicle crashes. With a growing population of older drivers in the United States, innovations in screening methods are needed to identify older adults at greatest risk of decline. This study used machine learning techniques to analyze real-world naturalistic driving data to identify depression status in older adults and examined whether specific demographics and medications improved model performance. We analyzed two years of GPS data from 157 older adults, including 81 with major depressive disorder, using XGBoost and logistic regression models. The top-performing model achieved an area under the curve of 0.86 with driving features combined with total medication use. These findings suggest that naturalistic driving data holds high potential as a functional digital neurobehavioral marker for AI identifying depression in older adults on a national scale, thereby ensuring equitable access to treatment.
Trustworthy and Responsible AI for Human-Centric Autonomous Decision-Making Systems
Dehghani F, Dibaji M, Anzum F, Dey L, Basdemir A, Bayat S, Boucher J-C, Drew S, Eaton SE, Frayne R, Ginde G, Harris A, Ioannou Y, Lebel C, Lysack J, Salgado Arzuaga L, Stanley E, Souza R, de Souza Santos R, Wells L, Williamson T, Wilms M, Wahid Z, Ungrin M, Gavrilova M, Bento M
Transactions on Machine Learning Research
Artificial Intelligence (AI) has paved the way for revolutionary decision-making processes, which if harnessed appropriately, can contribute to advancements in various sectors, from healthcare to economics. However, its black box nature presents significant ethical challenges related to bias and transparency. AI applications are hugely impacted by biases, presenting inconsistent and unreliable findings, leading to significant costs and consequences, highlighting and perpetuating inequalities and unequal access to resources. Hence, developing safe, reliable, ethical, and Trustworthy AI systems is essential.
Our team of researchers working with Trustworthy and Responsible AI, part of the Transdisciplinary Scholarship Initiative within the University of Calgary, conducts research on Trustworthy and Responsible AI, including fairness, bias mitigation, reproducibility, generalization, interpretability, and authenticity. In this paper, we review and discuss the intricacies of AI biases, definitions, methods of detection and mitigation, and metrics for evaluating bias. We also discuss open challenges with regard to the trustworthiness and widespread application of AI across diverse domains of human-centric decision making, as well as guidelines to foster Responsible and Trustworthy AI models.
LungXplain: Chunk-Level Modeling for Precise Detection and Localization of Respiratory Events
Pourebrahim T, Bayat S, Bento M, Curiel L
IEEE SENSORS Conference, Vancouver, Canada
This paper introduces LungXplain, a novel framework for detecting adventitious lung sounds that addresses the common challenges of coarse annotations and model inexplicability. Our core innovation is a fine-grained, chunk-level classification approach. We divide respiratory audio segments into short chunks, classify each using an XGBoost model trained on a comprehensive set of signal-based features, and then aggregate these predictions to determine the segment's overall label. Since each chunk inherits a potentially noisy label from its parent segment, this weakly supervised strategy allows our model to temporally localize pathological events like crackles and wheezes. On the official ICBHI 2017 test set, our method achieves a new state-of-the-art ICBHI Score of 95.71%, significantly outperforming previous benchmarks. The framework's explainability stems from its ability to visually pinpoint the exact temporal locations of detected pathological sounds, providing clear insight into the model's decision-making process. A key limitation, however, is that the practical utility of this visual explainability has not yet been validated in studies with medical experts.
Automated Helmet Detection in Construction Sites using UWB and Machine Learning
Shahbazi Ojghaz A, Bayat S, Sadeghpour F
International Symposium on Automation and Robotics in Construction, Montreal, Canada
Ensuring the proper use of personal protective equipment (PPE), particularly helmets, is crucial for enhancing safety on construction sites. This study proposes a novel approach for detecting helmet usage and worker states using ultra-wideband (UWB) localization sensors combined with machine learning algorithms. Unlike traditional sensor-based or image based systems, the proposed method integrates helmet detection into existing proximity warning systems, offering a cost-effective solution without the need for additional hardware. Data was collected from 12 participants in a controlled environment using UWB sensors, and a machine learning model was developed to classify four worker states: standing with a helmet, standing while holding a helmet, walking with a helmet, and walking while holding a helmet. The Gradient Boosting Decision Trees (GBDT) algorithm was selected for model development due to its superior performance. The model achieved an overall accuracy of 74.46% and performed well in detecting unsafe conditions. However, variability was observed in identifying safe states. Additionally, the study explored the impact of worker height on Z-axis localization error, revealing a correlation that suggests the need for height adjusted safety monitoring systems. This research demonstrates that UWB-based systems can enhance PPE monitoring, reduce computational costs, and address privacy concerns, while also highlighting areas for future improvement, such as expanding the model to detect other PPE and refining its ability to differentiate between worker postures.
Wrist-to-Back Signal Reconstruction for Real-World Gait Monitoring via Frequency-Domain Modelling
Aboudeif S, Alizadeh S, Bayat S
IEEE SENSORS Conference, Vancouver, Canada
Wearable accelerometers are commonly used for gait analysis in healthcare and rehabilitation, providing objective, continuous, real-world data. While lower-back sensors are widely used for capturing gait features, however, their long-term use is limited by discomfort and impracticality. Wrist-worn sensors are more user-friendly but less accurate for specific gait metrics. This study presents a novel, proof-ofconcept method to reconstruct lower-back accelerometer signals using only wrist-based data, via a frequency-domain transfer function approach. Using a two-day naturalistic, freeliving dataset from multiple participants wearing accelerometers on both the wrist and lower back, gait bouts were detected and clustered to extract representative gait cycles. A spectral transfer function modelled the wrist-lower back relationship. The proposed method achieved high intraindividual accuracy (Pearson correlation =0.9141; RMSE =0.4351) and maintained robust performance across individuals (mean Pearson correlation =0.827; RMSE =0.604), demonstrating its potential to generalize across demographic variability. Unlike prior deep learning-based models, our approach is interpretable, computationally efficient, and does not require large training datasets, making it well-suited for real-world applications. These findings highlight the feasibility of replacing Lower back sensors with wrist-worn devices, making it suitable for long-term, passive gait monitoring in clinical and consumer health applications.
Personalized POI Prediction from GPS Data Using Graph-Enhanced Transformers: Toward an Intelligent Navigation System
Wallich M, Babulal GM, Bayat S
IEEE SENSORS Conference, Vancouver, Canada
Advances in wearable sensor systems have enabled continuous monitoring of real-world human mobility, offering new opportunities for intelligent navigation support. This work presents a GPS-based spatial navigation framework that leverages next Point-of-Interest (POI) prediction to forecast future locations based on historical movement patterns. The model is pretrained on St. Louis county Foursquare check-in data and fine-tuned using a five-year longitudinal driving dataset from an older adult. By combining graph neural networks, temporal encodings, and attention mechanisms, our system captures individualized mobility behaviours and achieves high top-5 prediction accuracy. These results highlight the potential of sensor-driven, AI-powered systems for delivering personalized mobility support in aging and health contexts.
Navigating Fairness in Healthcare: A Comparative Analysis of Single-Stage and Multi-Stage Bias Mitigation Strategies
Dehghani F, Paiva P, Malik N, Lin J, Bayat S, Bento M
SANER-C, Montreal, Canada
Despite machine learning’s potential to revolutionize healthcare decision-making, systematic biases in algorithms and training data can worsen already-existing health inequities among different populations. Employing bias mitigation strategies can enhance the reliability and fairness of AI healthcare systems across diverse patient populations. The aim of this study is to develop a framework for evaluating various bias mitigation techniques and quantitatively measuring fairness in binary classification. We examined these techniques across pre-processing, in-processing, and post-processing stages, utilizing both single- and multi-stage intervention approaches. Our evaluation focused on their impact on model reliability, employing group and individual fairness metrics with a particular emphasis on the sensitive attribute of age. Our analysis focused on two healthcare datasets—stroke prediction and breast cancer diagnosis—highlighting the trade-offs between fairness and accuracy. Results show that Adversarial Debiasing significantly improves fairness by 95% in the breast cancer dataset without compromising accuracy. For stroke prediction, Reweighing is the most effective, enhancing fairness by 41% with minimal accuracy impact. These findings indicate that no single classifier or bias mitigation strategy fits all datasets; understanding context and experimenting with various methods is essential. While multistage bias mitigation can reduce bias, it may not always be effective, and efforts to minimize bias often come at the cost of accuracy, potentially hindering overall model performance in classification tasks.
Assessment of Anchor Configuration Effects on UWB Positioning Accuracy for Construction Safety Applications
Shahbazi Ojghaz A, Chang D, Shoemaker-Zuk R, Bayat S, Sadeghpour F
Canadian Society for Civil Engineering Conference, Niagara, Canada
Ultra-Wideband (UWB) Real-Time Location Systems (RTLS) have demonstrated significant potential for construction site safety applications due to their high accuracy and cost-effectiveness. However, the dynamic nature of construction environments often necessitates frequent reconfiguration of the UWB network, particularly the anchor positions. This study investigates the impact of anchor-related parameters on UWB RTLS positioning performance through systematic laboratory experiments. A precise measurement grid was established to evaluate the effects of anchor tilt, height, quantity, and multiple tag operations. The baseline configuration achieved mean accuracies of 5.23 cm and 19.95 cm for twodimensional and three-dimensional positioning, respectively. Results indicate that anchor height significantly influences positioning accuracy, with performance deteriorating by more than 50% in twodimensional and by a factor of five in three-dimensional measurements when anchors were positioned below tag height. Asymmetric anchor heights led to non-uniform accuracy degradation across the measurement space, particularly in regions near lower-height anchors. Contrary to expectations, increasing anchor quantity resulted in degraded performance, suggesting the importance of geometric optimization over anchor numbers. Multiple tag operations showed uniform but significant accuracy deterioration of 67.14% and 48.89% in two-dimensional and three-dimensional positioning, respectively. While these configuration changes led to decreased performance compared to the baseline, the system maintained sub-meter accuracy across all experimental configurations. These findings provide practical insights for optimal UWB RTLS deployment in construction environments, particularly regarding anchor placement strategies and network design considerations.
Impact of Cognitive Impairment on Driving Behaviour and Route Choices of Older Drivers: A Real-World Driving Study
Derafshi R, Babulal G, Bayat S
Scientific Reports
Maintaining driving independence is important for older adults. However, cognitive decline, a common issue in older populations, can impair older adults’ driving abilities and overall safety on the roads. This study explores how cognitive impairment influences driving patterns and driving choices among older adults. We analyzed real-world driving patterns of 246 older adults using GPS dataloggers. Our sample included 230 cognitively normal older adults (CN; Clinical Dementia Rating
[CDR] = 0) and 16 older adults with incident cognitive impairment (ICI; CDR = 0.5). The CN group had an average age of 68.2 years, with 46% females and an average of 16.5 years of education, while the ICI group’s average age was 69.2 years, with 36% females and an average of 16.0 years of education. We employed spatial clustering and hashing algorithms to evaluate driving behaviours. Significant differences emerged: The ICI group used fewer distinct routes to their most common destination. These differences can be leveraged to develop driving as a digital biomarker for the early detection and continuous monitoring of cognitive impairment.
Alzheimer Disease Detection Studies: Perspective on Multi-Modal Data
Dehghani F, Derafshi R, Lin J, Bayat S, Bento M
Yearbook of Medical Informatics
Objectives: Alzheimer's Disease (AD) is one of the most common neurodegenerative diseases, resulting in progressive cognitive decline, and so accurate and timely AD diagnosis is of critical importance. To this end, various medical technologies and computer-aided diagnosis (CAD), ranging from biosensors and raw signals to medical imaging, have been used to provide information about the state of AD. In this survey, we aim to provide a review on CAD systems for automated AD detection, focusing on different data types: namely, signals and sensors, medical imaging, and electronic medical records (EMR).
Methods: We explored the literature on automated AD detection from 2022-2023. Specifically, we focused on various data resources and reviewed several preprocessing and learning methodologies applied to each data type, as well as evaluation metrics for model performance evaluation. Further, we focused on challenges, future perspectives, and recommendations regarding automated AD diagnosis.
Results: Compared to other modalities, medical imaging was the most common data type. The prominent modality was Magnetic Resonance Imaging (MRI). In contrast, studies based on EMR data type were marginal because EMR is mostly used for AD prediction rather than detection. Several challenges were identified: data scarcity and bias, imbalanced datasets, missing information, anonymization, lack of standardization, and explainability.
Conclusion: Despite recent developments in automated AD detection, improving the trustworthiness and performance of these models, and combining different data types will improve usability and reliability of CAD tools for early AD detection in the clinical practice.
Fairness in Healthcare: Assessing Data Bias and Algorithmic Fairness
Dehghani F, Malik N, Lin J, Bayat S, Bento M
SIPAIM, Antigua, Guatemala
With the wide employment of Artificial Intelligence-based systems in healthcare, there is an increasing need to further quantify and analyze potential biases in data and algorithms, mitigating health disparities. We aim to investigate publicly available healthcare datasets to identify potential biases, and investigate the impact of these biases on the predictive performance of the models. The results indicate that bias in these datasets exerts a negative impact on the fairness of the learned models. In the context of a highly imbalanced dataset, we further investigate the impact of various imbalance handling techniques on both the performance and fairness of the models. This study underscores the importance of early detection of bias to mitigate the risk of introducing such biases in real-world applications, particularly within a sensitive domain, such as healthcare.
Edge Computing for Real-time Monitoring Systems in Construction
Shahbazi Ojghaz A, Sadeghpour F, Bayat S
Canadian Society for Civil Engineering Conference, Niagara, Canada
This paper presents an innovative approach to real-time monitoring in construction management by integrating Building Information Modeling (BIM) with edge computing to develop a hybrid system. The study explores how this proposed system addresses the critical demand for immediate data access and rapid response, overcoming the limitations of traditional standalone and client–server systems in dynamic construction environments. The research follows the development and performance assessment of a prototype through a case study focused on site safety monitoring. The prototype utilizes ultra-wideband (UWB) sensor technology for precise worker tracking and alerting within hazard zones, offering significant improvements in time latency compared to conventional systems. Results indicate that the hybrid system substantially reduces the response time to potential hazards, enhancing safety measures. The hybrid system's architecture also allows for standardized and flexible data management with BIM, facilitating the expansion of applications without complex modifications to server architecture. Through rigorous experimentation, the study demonstrates the hybrid system's superior performance with the ability to issue alerts more rapidly and reliably than client–server models. The research underscores the potential of this hybrid approach to advance real-time monitoring in construction, promising benefits for safety, efficiency, and management, while also acknowledging the limitations of hardware capabilities and real-world construction conditions.
Researched Apps Used in Dementia Care for People Living with Dementia and Their Informal Caregivers: Systematic Review on App Features, Security, and Usability
Ye B, Chu CH, Bayat S, Babineau J, How TV, Mihailidis A
Journal of Medical Internet Research
Background: Studies have shown that mobile apps have the potential to serve as nonpharmacological interventions for dementia care, improving the quality of life of people living with dementia and their informal caregivers. However, little is known about the needs for and privacy aspects of these mobile apps in dementia care.
Objective: This review seeks to understand the landscape of existing mobile apps in dementia care for people living with dementia and their caregivers with respect to app features, usability testing, privacy, and security.
Methods: ACM Digital Library, Cochrane Central Register of Controlled Trials, Compendex, Embase, Inspec, Ovid MEDLINE, PsycINFO, and Scopus were searched. Studies were included if they included people with dementia living in the community, their informal caregivers, or both; focused on apps in dementia care using smartphones or tablet computers; and covered usability evaluation of the app. Records were independently screened, and 2 reviewers extracted the data. The Centre for Evidence-Based Medicine critical appraisal tool and Mixed Methods Appraisal Tool were used to assess the risk of bias in the included studies. Thematic synthesis was used, and the findings were summarized and tabulated based on each research aim.
Results: Overall, 44 studies were included in this review, with 39 (89%) published after 2015. In total, 50 apps were included in the study, with more apps developed for people living with dementia as end users compared with caregivers. Most studies (27/44, 61%) used tablet computers. The most common app feature was cognitive stimulation. This review presented 9 app usability themes: user interface, physical considerations, screen size, interaction challenges, meeting user needs, lack of self-awareness of app needs, stigma, technological inexperience, and technical support. In total, 5 methods (questionnaires, interviews, observations, logging, and focus groups) were used to evaluate usability. There was little focus on the privacy and security aspects, including data transfer and protection, of mobile apps for people living with dementia.
Conclusions: The limitations of this review include 1 reviewer conducting the full-text screening, its restriction to studies published in English, and the exclusion of apps that lacked empirical usability testing. As a result, there may be an incomplete representation of the available apps in the field of dementia care. However, this review highlights significant concerns related to the usability, privacy, and security of existing mobile apps for people living with dementia and their caregivers. The findings of this review provide a valuable framework to guide app developers and researchers in the areas of privacy policy development, app development strategies, and the importance of conducting thorough usability testing for their apps. By considering these factors, future work in this field can be advanced to enhance the quality and effectiveness of dementia care apps.
Associations Between Plasma, Imaging, and Cerebrospinal Fluid Biomarkers with Driving Behavior and Cognitive Tests: Implications for Biomarker Usefulness
Roe CM, Bayat S, Babulal GM
Journal of Alzheimer's Disease Reports
Background: Declines in instrumental activities of daily living like driving are hallmarks sequelae of Alzheimer's disease (AD). Although driving has been shown to be associated with traditional imaging and cerebrospinal fluid (CSF) biomarkers, it is possible that some biomarkers have stronger associations with specific aspects of driving behavior. Furthermore, associations between newer plasma biomarkers and driving behaviors are unknown.
Objective: This study assessed the extent to which individual plasma, imaging, and CSF biomarkers are related to specific driving behaviors and cognitive functions among cognitively normal older adults.
Methods: We analyzed naturalistic driving behavior from cognitively healthy older drivers (N = 167, 47% female, mean age = 73.3 years). All participants had driving, clinical, and demographic data and completed biomarker testing, including imaging, CSF, and/or plasma, within two years of study commencement.
Results: AD biomarkers were associated with different characteristics of driving and cognitive functioning within the same individuals. Elevated levels of plasma Aβ40 were associated with more speeding incidents, higher levels of CSF tau were related to shorter duration of trips, and higher CSF neurofilament light chain values were associated with traveling shorter distances, smaller radius of gyration, and fewer trips at night. We demonstrated that plasma, like CSF and imaging biomarkers, were helpful in predicting everyday driving behaviors.
Conclusions: These findings suggest that different biomarkers offer complementary information with respect to driving behaviors. These distinct relationships may help in understanding how different biological changes that occur during the preclinical stage of AD can impact various sensorimotor and cognitive processes.
Everyday Driving and Plasma Biomarkers in Alzheimer's Disease: Leveraging Artificial Intelligence to Expand Our Diagnostic Toolkit
Bayat S, Roe CM, Schindler S, Murphy SA, Doherty JM, Johnson AM, Walker A, Ances BM, Morris JC, Babulal GM
Journal of Alzheimer's Disease
Background: Driving behavior as a digital marker and recent developments in blood-based biomarkers show promise as a widespread solution for the early identification of Alzheimer's disease (AD).
Objective: This study used artificial intelligence methods to evaluate the association between naturalistic driving behavior and blood-based biomarkers of AD.
Methods: We employed an artificial neural network (ANN) to examine the relationship between everyday driving behavior and plasma biomarker of AD. The primary outcome was plasma Aβ42/Aβ40, where Aβ42/Aβ40 0.051].
Conclusion: Blood-based AD biomarkers offer a novel opportunity to establish the efficacy of naturalistic driving as an accessible digital marker for AD pathology in driving research.
Adverse driving behaviors increase over time as a function of preclinical Alzheimer's disease biomarkers
Doherty JM, Murphy SA, Bayat S, Wisch JK, Johnson AM, Walker A, Schindler SE, Ances BM, Morris JC, Babulal GM
Alzheimer's & Dementia
Introduction: We investigated the relationship between preclinical Alzheimer's disease (AD) biomarkers and adverse driving behaviors in a longitudinal analysis of naturalistic driving data.
Methods: Naturalistic driving data collected using in‐vehicle dataloggers from 137 community‐dwelling older adults (65+) were used to model driving behavior over time. Cerebrospinal fluid (CSF) biomarkers were used to identify individuals with preclinical AD. Additionally, hippocampal volume and cognitive biomarkers for AD were investigated in exploratory analyses.
Results: CSF biomarkers predicted the longitudinal trajectory of the incidence of adverse driving behavior. Abnormal amyloid beta (Aβ42/Aβ40) ratio was associated with an increase in adverse driving behaviors over time compared to ratios in the normal/lower range.
Discussion: Preclinical AD is associated with increased adverse driving behavior over time that cannot be explained by cognitive changes. Driving behavior as a functional, neurobehavioral marker may serve as an early detection for decline in preclinical AD. Screening may also help prolong safe driving as older drivers age.
Driving assessment in preclinical Alzheimer's disease: progress to date and the path forward
Bayat S, Roe CM
Alzheimer's Research & Therapy
Background: Changes in driving behaviour may start at the preclinical stage of Alzheimer’s disease (AD), where the underlying AD biological process has begun in the presence of cognitive normality. Here, we summarize the emerging evidence suggesting that preclinical AD may impact everyday driving behaviour.
Main: Increasing evidence links driving performance and behaviour with AD biomarkers in cognitively intact older adults. These studies have found subtle yet detectable differences in driving associated with AD biomarker status among cognitively intact older adults.
Conclusion: Recent studies suggest that changes in driving, a highly complex activity, are linked to, and can indicate the presence of, neuropathological AD. Future research must now examine the internal and external validity of driving for widespread use in identifying biological AD.
An event-based model and a map visualization approach for spatiotemporal association relations discovery of diseases diffusion
Habibi R, Alesheikh AA, Bayat S
Sustainable Cities and Society
Infectious disease diffusion is inherently a complex spatiotemporal phenomenon. Simplifying this complexity to discover the associated structure of the city is of great importance. However, existing approaches mainly focus on distance property in geographic space to examine randomness, dispersion, or clustered structure of the disease distribution. While, the outbreak continuously changes its properties, shapes, or locations. Regardless of this adjacency-based structure, there may be associated spatial units that exhibit similar behaviors towards the outbreak fluctuations in a city. To reveal these characteristics, this research proposes a novel event-based spatiotemporal model, mining associated areas in space and time simultaneously. This model was applied to the cases rate of COVID-19 at the ZIP Code level in New York City. The results showed that the proposed approach could sufficiently address the spatiotemporal association relationships. To better understand the discovered associations, a map visualization approach is introduced, allowing recognition of these association relations at a glance. This approach develops a deep understanding of the spatiotemporal structure of the outbreak and better manifests the association and cause-and-effect relations between ZIP Code areas. The results provide good assets for the construction of healthy resilient cities with the function of preventing epidemic crises in the future.
Neuropsychological Correlates of Changes in Driving Behavior Among Clinically Healthy Older Adults
Aschenbrenner AJ, Murphy SA, Doherty JM, Johnson AM, Bayat S, Walker A, Peña Y, Hassenstab J, Morris JC, Babulal GM
The Journals of Gerontology: Series B
Objectives: To determine the extent to which cognitive domain scores moderate change in driving behavior in cognitively healthy older adults using naturalistic (Global Positioning System-based) driving outcomes and to compare against self-reported outcomes using an established driving questionnaire.
Methods: We analyzed longitudinal naturalistic driving behavior from a sample (N = 161, 45% female, mean age = 74.7 years, mean education = 16.5 years) of cognitively healthy, nondemented older adults. Composite driving variables were formed that indexed “driving space” and “driving performance.” All participants completed a baseline comprehensive cognitive assessment that measured multiple domains as well as an annual self-reported driving outcomes questionnaire.
Results: Across an average of 24 months of naturalistic driving, our results showed that attentional control, broadly defined as the ability to focus on relevant aspects of the environment and ignore distracting or competing information as measured behaviorally with tasks such as the Stroop color naming test, moderated change in driving space scores over time. Specifically, individuals with lower attentional control scores drove fewer trips per month, drove less at night, visited fewer unique locations, and drove in smaller spaces than those with higher attentional control scores. No cognitive domain predicted driving performance such as hard braking or sudden acceleration.
Discussion: Attentional control is a key moderator of change over time in driving space but not driving performance in older adults. We speculate on mechanisms that may relate attentional control ability to modifications of driving behaviors.
Driving, Social Distancing, Protective, and Coping Behaviors of Older Adults Before and During COVID-19
Roe CM, Bayat S, Hicks J, Johnson AM, Murphy S, Doherty JM, Babulal G
Journal of Applied Gerontology
A thorough understanding of individual characteristics of older adults during the COVID-19 pandemic is critical for managing the ongoing pandemic course and planning for the future pandemics. Here, we explore the impact of the COVID-19 pandemic on driving, social distancing, protective, and coping behaviors of older adults. This study reports data on participants aged above 65 whose driving behaviors are being monitored using Global Positioning System (GPS) devices. Participants completed a COVID-19 survey in May 2020. We found that older adults decreased their number of days driving, number of trips per day, as well as average driving speed, and had fewer speeding incidents following COVID-19 onset. We also show that female and African American older adults engaged in more positive coping and cleaning behaviors, and had greater decreases in the number of days driving during the pandemic. The findings highlight the importance of considering older adults’ individual characteristics for an equitable response to the COVID-19 pandemic.
A GPS-Based Framework for Understanding Outdoor Mobility Patterns of Older Adults with Dementia: An Exploratory Study
Bayat S, Naglie G, Rapoport MJ, Stasiulis E, Widener M, Mihailidis A
Gerontology
Introduction: An active lifestyle may protect older adults from cognitive decline. Yet, due to the complex nature of outdoor environments, many people living with dementia experience decreased access to outdoor activities. In this context, conceptualizing and measuring outdoor mobility is of great significance. Using the global positioning system (GPS) provides an avenue for capturing the multi-dimensional nature of outdoor mobility. The objective of this study is to develop a comprehensive framework for comparing outdoor mobility patterns of cognitively intact older adults and older adults with dementia using passively collected GPS data.
Methods: A total of 7 people with dementia (PwD) and 8 cognitively intact controls (CTLs), aged 65 years or older, carried a GPS device when travelling outside their homes for 4 weeks. We applied a framework incorporating 12 GPS-based indicators to capture spatial, temporal, and semantic dimensions of outdoor mobility.
Results: Despite a small sample size, the application of our mobility framework identified several significant differences between the 2 groups. We found that PwD participated in more medical-related (Cliff's Delta = 0.71, 95% CI: 0.34-1) and fewer sport-related (Cliff's Delta = -0.78, 95% CI: -1 to -0.32) activities compared to the cognitively intact CTLs. Our results also suggested that longer duration of daily walking time (Cliff's Delta = 0.71, 95% CI: 0.148-1) and longer outdoor activities at night, after 8 p.m. (Hedges' g = 1.42, 95% CI: 0.85-1.09), are associated with cognitively intact individuals.
Conclusion: Based on the proposed framework incorporating 12 GPS-based indicators, we were able to identify several differences in outdoor mobility in PwD compared with cognitively intact CTLs.
GPS driving: a digital biomarker for preclinical Alzheimer disease
Bayat S, Babulal G, Schindler SE, Fagan AM, Morris JC, Mihailidis A, Roe CM
Alzheimer's Research & Therapy
Background: Alzheimer disease (AD) is the most common cause of dementia. Preclinical AD is the period during which early AD brain changes are present but cognitive symptoms have not yet manifest. The presence of AD brain changes can be ascertained by molecular biomarkers obtained via imaging and lumbar puncture. However, the use of these methods is limited by cost, acceptability, and availability. The preclinical stage of AD may have a subtle functional signature, which can impact complex behaviours such as driving. The objective of the present study was to evaluate the ability of in-vehicle GPS data loggers to distinguish cognitively normal older drivers with preclinical AD from those without preclinical AD using machine learning methods.
Methods: We followed naturalistic driving in cognitively normal older drivers for 1 year with a commercial in-vehicle GPS data logger. The cohort included n = 64 individuals with and n = 75 without preclinical AD, as determined by cerebrospinal fluid biomarkers. Four Random Forest (RF) models were trained to detect preclinical AD. RF Gini index was used to identify the strongest predictors of preclinical AD.
Results: The F1 score of the RF models for identifying preclinical AD was 0.85 using APOE ε4 status and age only, 0.82 using GPS-based driving indicators only, 0.88 using age and driving indicators, and 0.91 using age, APOE ε4 status, and driving. The area under the receiver operating curve for the final model was 0.96.
Conclusion: The findings suggest that GPS driving may serve as an effective and accurate digital biomarker for identifying preclinical AD among older adults.
Bringing the 'Place' to Life Space in Gerontology Research
Bayat S, Widener M, Mihailidis A
Gerontology
Understanding older adults' relationships with their environments and the way this relationship evolves over time have been increasingly acknowledged in gerontological research. This relationship is often measured in terms of life-space, defined as the spatial area through which a person moves within a specific period of time. Life-space is traditionally reported using questionnaires or travel diaries and is, thus, subject to inaccuracies. More recently, studies are using a global positioning system to accurately measure life-space. Although life-space provides useful insights into older adults' relationships with their environment, it does not capture the inherent complexities of environmental exposures. In the fields of travel behaviour and health geography, a substantial amount of research has looked at people's spatial behaviour using the notion of "Activity Space," allowing for increasing sophistication in understanding older adults' experience of their environment. This manuscript discusses developments and directions for extending the life-space framework in environmental gerontology by drawing on the advancements in the activity space framework.
Outdoor life in dementia: How predictable are people with dementia in their mobility?
Bayat S, Mihailidis A
Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring
Introduction: People with dementia (PWD) often become disoriented, which increases their risk of getting lost. This article explores the extent to which we can predict future whereabouts of PWD by learning from their past mobility patterns using Global Positioning System (GPS) tracking devices.
Methods: Seven older adults with dementia and eight healthy older adults completed 8 weeks of GPS data collection. We computed the probability that an appropriate algorithm can correctly predict the participant's future destinations using spatial and temporal patterns in each participant's GPS trajectories.
Results: Relying on both spatial and temporal patterns, our results suggest that a 4-week record of mobility patterns displays 95% potential predictability across the dementia group, which is significantly higher than 92% potential predictability among the controls, t(13) = -3.39, P
Inferring Destinations and Activity Types of Older Adults from GPS Data: Algorithm Development and Validation
Bayat S, Naglie G, Rapoport MJ, Stasiulis E, Chikhaoui B, Mihailidis A
JMIR Aging
Background: Outdoor mobility is an important aspect of older adults' functional status. GPS has been used to create indicators reflecting the spatiotemporal dimensions of outdoor mobility for applications in health and aging. However, outdoor mobility is a multidimensional construct. There is, as of yet, no classification algorithm that groups and characterizes older adults' outdoor mobility based on its semantic aspects (ie, mobility intentions and motivations) by integrating geographic and domain knowledge.
Objective: This study assesses the feasibility of using GPS to determine semantic dimensions of older adults' outdoor mobility, including destinations and activity types.
Methods: A total of 5 healthy individuals, aged 65 years or older, carried a GPS device when traveling outside their homes for 4 weeks. The participants were also given a travel diary to record details of all excursions from their homes, including date, time, and destination information. We first designed and implemented an algorithm to extract destinations and infer activity types (eg, food, shopping, and sport) from the GPS data. We then evaluated the performance of the GPS-derived destination and activity information against the traditional diary method.
Results: Our results detected the stop locations of older adults from their GPS data with an F1 score of 87%. On average, the extracted home locations were within a 40.18-meter (SD 1.18) distance of the actual home locations. For the activity-inference algorithm, our results reached an F1 score of 86% for all participants, suggesting a reasonable accuracy against the travel diary recordings. Our results also suggest that the activity inference's accuracy measure differed by neighborhood characteristics (ie, Walk Score).
Conclusions: We conclude that GPS technology is accurate for determining semantic dimensions of outdoor mobility. However, further improvements may be needed to develop a robust application of this system that can be adopted in clinical practice.
bottom of page