Canada

Cardiovascular diseases remain the leading global cause of death, with aging populations facing disproportionate burden. While automated ECG interpretation using convolutional neural networks (CNNs) shows promise for democratizing cardiac diagnostics, age- related performance bias has not been systematically evaluated. We developed a low-resource CNN-based system using continuous wavelet transform (CWT) scalograms from single-lead ECG signals (Lead II) for three-class rhythm classification: Sinus Rhythm (SR), Sinus Bradycardia (SB), and Atrial Fibrillation/Flutter (AFIB). The transfer learning approach adapted AlexNet architecture trained on 45,152 expert-annotated 12-lead ECGs from the Chapman University database. Age-stratified evaluation was conducted across three demographic groups: young adults (0-40 years), middle-aged (40-65 years), and elderly (65+ years). On the test dataset, the model achieved 96.88% overall accuracy, outperforming existing literature on multi-class arrhythmia classification. However, significant age-related bias was identified, particularly affecting elderly patients. AFIB classification precision declined by 4 percentage points from middle aged (98.5%) to elderly patients (94.5%), while SB classification demonstrated remarkable robustness across all age groups (&gt;99.2% specificity). Overall accuracy decreased from 96.9% to 96.0% between young and elderly cohorts. This study represents the first comprehensive age- bias analysis in CNN-based ECG classification, revealing systematic performance degradation in elderly populations. The findings have critical implications for healthcare equity and highlight the need for age-aware training strategies to ensure equitable deployment of automated cardiac diagnostics in aging global populations.

Aging Powerfully: AI Blueprint for Chronic Disease Management and Social Determinants of Longevity

Age Group Robustness in CNN-Based ECG Rhythm Classification Using Wavelet Scalograms

healthcare equity

time-series feature map

continuous wavelet transform (cwt)

arrhythmia

convolutional neural network (cnn)

age bias

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.<br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

AAAI 2026

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

Cardiovascular diseases remain the leading global cause of death, with aging populations facing disproportionate burden. While automated ECG interpretation using convolutional neural networks (CNNs) shows promise for democratizing cardiac diagnostics, age- related performance bias has not been systematically evaluated. We developed a low-resource CNN-based system using continuous wavelet transform (CWT) scalograms from single-lead ECG signals (Lead II) for three-class rhythm classification: Sinus Rhythm (SR), Sinus Bradycardia (SB), and Atrial Fibrillation/Flutter (AFIB). The transfer learning approach adapted AlexNet architecture trained on 45,152 expert-annotated 12-lead ECGs from the Chapman University database. Age-stratified evaluation was conducted across three demographic groups: young adults (0-40 years), middle-aged (40-65 years), and elderly (65+ years). On the test dataset, the model achieved 96.88% overall accuracy, outperforming existing literature on multi-class arrhythmia classification. However, significant age-related bias was identified, particularly affecting elderly patients. AFIB classification precision declined by 4 percentage points from middle aged (98.5%) to elderly patients (94.5%), while SB classification demonstrated remarkable robustness across all age groups (>99.2% specificity). Overall accuracy decreased from 96.9% to 96.0% between young and elderly cohorts. This study represents the first comprehensive age- bias analysis in CNN-based ECG classification, revealing systematic performance degradation in elderly populations. The findings have critical implications for healthcare equity and highlight the need for age-aware training strategies to ensure equitable deployment of automated cardiac diagnostics in aging global populations.

technical paper

2025 National Inter-University Health Data & AI Inquiry Program - this program is a unique project-based inquiry and science communication program that brings together students from diverse academic backgrounds and research interests. Presented research projects explore how generative AI is being applied to medication management, mental health screening, and chronic disease monitoring to align with the future of pharmacy practice.

2025 National Inter-University Health Data & AI Inquiry Program is a unique project-based inquiry and science communication program that brings together students from diverse academic backgrounds and research interests. 

Osteoporosis is a slow, often undetected condition that leads to decreased bone density and increased fracture risk, particularly among the elderly. Despite its prevalence, most people are unaware of its risk until the condition gets severe, making early detection both critical and challenging. This project supports early screening and personalized risk communication through artificial intelligence (AI) and natural language processing (NLP) to develop a method that predicts the likelihood of osteoporosis in a patient from their health records and explains the associated risks in simple and understandable language. Due to the scarcity of reliable publicly available datasets, a synthetic dataset was created based on a medically reviewed questionnaire having 50 questions varying by gender and including both binary, numerical, and alphanumerical inputs. Containing the key clinical attributes such as bone mineral density, age, symptom history, sex, calcium/vitamin D levels, and lifestyle factors. The classification pipeline utilizes transformer-based models like DistilBERT, and RoBERTa to assess osteoporosis risk, while LLaMA 3 8B, LLaMA 3.1 405B parameters, and DeepSeek LLM 7B are employed for natural language summarization and explanation. By integrating advanced machine learning with NLP, this work offers a dual benefit. Model performance is evaluated using ROUGE score for summarization accuracy, and perplexity score for language fluency. It's designed as a universal tool that everyone can use. The main goal is contributing to earlier diagnosis, improving public health, and longevity-focused healthcare by taking precautionary measures as soon as possible.

AI-Based Early Osteoporosis Risk Detection Using Questionnaire Responses

Alzheimer’s disease (AD) is the most common cause of dementia in older adults, posing a major public health and economic burden worldwide. While early detection of AD is critical to improving patient outcomes and delaying disease progression, existing diagnostic methods, such as cognitive testing, spinal fluid assays, and advanced brain imaging, are often costly, invasive, or inaccessible in community settings. Our project explores the role of artificial intelligence (AI) to enable scalable, non-invasive screening by jointly analyzing retinal images and electroencephalography (EEG) recordings. The retina, as an extension of the central nervous system, reflects structural changes such as nerve fiber thinning and microvascular abnormalities, while EEG captures functional brain activity that can reveal disruptions in neural connectivity and cognitive processing. Our research involves implementing a reproducible Jupyter- Notebook pipeline: retinal optical coherence tomography (OCT) scans are normalized via the min-max scaling technique, fed into a convolutional neural network (CNN) pretrained on large public image collections, then fine-tuned on expert-labeled fundus photographs to focus on disease-relevant features. Simultaneously, EEG signals undergo band-pass and notch filtering, segmentation into fixed-length windows, and z-score standardization via TensorFlow’s data before being classified by a compact EEGNet architecture. By integrating structural and functional outputs, our multimodal model substantially outperforms single-modality approaches, demonstrating enhanced sensitivity in detecting early AD and mild cognitive impairment. This research has the potential to inform the development of affordable, accessible diagnostic tools that can be deployed in clinical and community settings, particularly for early intervention in aging populations at risk of dementia.

AI-Driven Diagnosis and Early Detection of Alzheimer's Disease Using Retinal Imaging and EEG

Cardiovascular diseases (CVDs) are a leading cause of death in older adults. Early and accurate diagnosis is vital but challenging due to complex clinical and physiological data. This study presents an AI framework using knowledge graphs and graph neural networks (GNNs) to improve CVD detection in elderly patients. We constructed feature-based knowledge graphs from three clinical datasets and modeled ECG signals as graphs using the PTB-XL dataset. Multiple GNN architectures—including GCN, GraphSAGE, GAT, and RGCN—were trained and compared. Our results show that GCN achieves the highest accuracy (ROC AUC: 0.78) for classifying CVD risk. The approach demonstrates the value of integrating graph-based methods for early, interpretable, and accurate CVD prediction in older adults.

AI-Driven Early Detection of Cardiovascular Diseases Using Knowledge Graphs and GNNs for Elderly

Polycystic Ovary Syndrome (PCOS) is a common, multifactorial endocrine disorder affecting 8–13% of women of reproductive age. It is associated with chronic metabolic, cardiovascular, and reproductive complications that contribute to accelerated biological aging. Due to its heterogeneity, standard diagnostic and treatment approaches often fail to capture the diverse pathophysiological mechanisms involved. To address this, we developed an integrative artificial intelligence (AI) framework to stratify PCOS patients into clinically meaningful subgroups and predict individualized treatments based on genetic, hormonal, and metabolic profiles. Clinical and biochemical data were sourced from the NHANES database, while SNP associations were derived from the GWAS Catalog and dbSNP. Pathway enrichment was conducted using Enrichr. Unsupervised clustering identified three phenotypic subtypes linked to inflammatory, metabolic, and neuroendocrine features, each with distinct pathway enrichments such as NF-κB signaling, nitric oxide regulation, and synaptic transmission. Supervised machine learning models, including Random Forest and XGBoost, were used to predict optimal treatments—such as metformin for insulin resistance and oral contraceptives for hyperandrogenism—with high performance (accuracy = 0.99, precision = 0.95, recall = 0.96). SHAP analysis confirmed biological interpretability, highlighting fasting glucose, insulin, BMI, and androgen biomarkers as key predictors. This study demonstrates the utility of AI-based stratification and treatment matching to enhance PCOS management and promote healthier aging in women.

AI-Driven Multi-Omics Approach for PCOS Subtyping and Drug Prediction to Advance Precision Women’s Healthy Aging

Falls present significant health consequences for older adults, increasing the risks of severe injury, hospitalization, and loss of independence. In recent years, artificial intelligence (AI) powered wearable devices have emerged as a promising fall detection and prevention technology, particularly for older adults living with chronic mobility impairments. The objective of this narrative review is to synthesize current evidence on AI-powered wearable systems for the detection and prevention of falls, with a focus on performance in real-world settings. PubMed, Google Scholar, and Omni were searched for studies published between 2018 and 2025. A total of 17 peer-reviewed articles were included in this review, meeting the following inclusion criteria: (1) study of adults 55 years and older, (2) utilization of AI tools with potential for real-world clinical translation, and (3) reporting of patient-relevant outcomes. The majority of evaluated technologies integrated inertial measurement units (IMU) and deep learning models (e.g., convolutional neural networks [CNN], long short-term memory [LSTM], or hybrid models). Edge computing based and multimodal sensor integrated systems reported the best performance with real-time results. Systems tested in realistic environments produced the strongest results. Commercially available products were notably poorer in performance in terms of sensitivity. Furthermore, several investigated studies did not account for the impacts of transparency and fairness. In conclusion, wearable technologies powered by AI show promise for fall detection and prevention for older adults, particularly in real-world environments. Prioritizing accessibility, equity, and incorporation into geriatric care should be the focus of future studies.

AI-Powered Wearable Sensors for Fall Detection in Older Adults

Alzheimer’s Disease (AD), a progressive neurodegenerative disorder is characterized by β-amyloid plaques and tau neurofibrillary tangles, leading to neuroinflammation and neurodegeneration. Despite efforts to characterize and understand the pathogenesis AD, its etiology has not been well understood. This study aims to investigate the association between bacterial gut microbiota genus and AD outcomes. We compiled 327 samples (216 AD cases, 111 controls) from the GMrepo database. Using logistic regression, we analyzed the association between relative abundance of bacterial genera, adjusting for age and sex, to identify its link to AD. Results reveal distinct microbiome profiles in AD patients. We found higher abundances of Roseburia, Dorea, Ruminococcus, Streptococcus, Bifidobacterium, Akkermansia, Holdemania, Bilophila, Prevotella, and Collinsella in AD patients than controls. AD patients also have lower relative abundance of Bacteroides, Faecalibacterium, Alistipes, Parabacteroides, Coprococcus, and Dialister compared to healthy controls. In our regression model, Gemmingers and Bifidobacterium, Lactobacillus, and Ruminococcus were associated with increased odds of AD. On the other hand, Bacteroides, Coprococcus, and Alistipes were associated with reduced odds of AD. These findings suggest that gut microbiota may influence AD through neuroinflammation via the gut-brain axis. This study also provides evidence and foundation for future research into potentially utilizing microbiome-targeted or preventative intervention for AD management while emphasizing the need for large-scale longitudinal studies to better understand the dynamic changes of the gut microbiome and its mechanism to AD.

Association Between Gut-Microbiome and Alzheimer’s Disease: a Regression Analysis

Parkinson’s disease (PD) is a progressive neurodegenerative disorder that primarily affects aging populations. In Canada, PD impacts approximately 1 in every 500 people. The hallmark motor symptoms of PD include tremors, muscle rigidity, bradykinesia, and postural instability, which increase the risk of falls. As the disease progresses, patients often experience a loss of mobility, independence, and overall quality of life, highlighting the urgent need for more efficient, innovative, and patient- centered approaches to care. One promising avenue is the application of artificial intelligence in PD management, particularly through the development of smart drug delivery systems. These technologies aim to modulate medication release in real time based on an individual’s fluctuating symptoms.This study explores supervised and unsupervised machine learning techniques for the classification of Parkinson’s disease and the profiling of its motor symptoms using wrist-worn accelerometer data from the Parkinson’s Apple Watch Dataset. We first replicated a published supervised classification pipeline using the Bag-of-Symbolic-Fourier-Approximation-Symbols algorithm combined with Support Vector Machines. Our results closely matched the original study’s performance, validating the reliability and reproducibility of this approach for distinguishing PD from healthy controls and differential diagnoses.For unsupervised analysis, we extracted tremor-related features from filtered acceleration signals, including mean acceleration, amplitude envelope, and power spectral density metrics within the 3–12 Hz tremor band. Gaussian Mixture Models were used to identify latent tremor patterns, with the optimal number of clusters selected via the Bayesian Information Criterion. Dimensionality reduction using t-distributed Stochastic Neighbor Embedding enabled visual inspection of the resulting clusters.

Beyond Diagnosis: Enhancing Parkinson’s Disease Detection and Symptom Profiling Using Wearable Data

Dementia is largely underdiagnosed in early stages because cognitive and behavioural symptoms are often misattributed to aging. With global dementia cases on the rise, artificial intelligence offers a promising solution to support early detection. This study assesses whether easy-to-interpret white-box models can achieve diagnostic accuracy comparable to that of complex black-box models in dementia screening, striving to increase transparency without compromising predictive accuracy. With a sample of 358 adults from the OASIS-1 and OASIS-2 datasets, we trained white-box and black-box machine learning models on six clinical and demographic predictors of dementia: Mini Mental State Examination (MMSE), age, sex, education, socioeconomic score, and normalized whole-brain volume (nWBV). The Black-box models included Random Forest, XGBoost, and k-Nearest Neighbours, and the white-box model used LASSO-regularized logistic regression. AI Models were evaluated using accuracy, sensitivity, specificity, positive predictive values, and negative predictive values for both binary and multiclass classification tasks. In general, black-box models performed slightly better; however, the white-box model achieved a notably strong binary sensitivity. This suggests that white-box models were particularly effective in identifying individuals with cognitive impairment based on their clinical and demographic characteristics. While white-box models did not perform as strongly as black-box models, especially in terms of multiclass accuracy, they may still be helpful in healthcare, as their transparency enables physicians to interpret and verify predictions. The study's findings emphasize the trade-off between performance and interpretability.

Comparing White-Box Versus Black-Box Predictive AI Models for Dementia Using OASIS Data

This research leverages deep learning techniques to investigate the connectivity between social determinants of health (SDoH) and chronic disease-related mortality and life expectancy globally. We integrated multiple datasets, including World Bank socioeconomic indicators, WHO global health estimates, and income inequality measures, spanning from 2000 to 2021. Using exploratory data analysis, we identified global trends and significant correlations between five chronic disease outcomes, Alzheimer’s, asthma, cardiovascular diseases, diabetes, and kidney diseases, and various socioeconomic factors. We implemented multiple predictive models, including linear regression, XGBoost, and recurrent neural networks, including Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM), validated using 3-fold cross- validation. The parameters are selected using hyperparameter tuning techniques. XGBoost consistently outperformed other methods across diseases, achieving superior accuracy as measured by mean squared error (MSE) and coefficient of determination (R²). To ensure interpretability, SHAP (SHapley Additive exPlanations) values were applied, highlighting influential socioeconomic factors like income inequality, diabetes prevalence, and life expectancy. Our analysis revealed distinct patterns based on income groups and geographic regions, emphasizing the necessity of specific public health strategies. This research uniquely contributes by integrating interpretable AI techniques, providing actionable insights into the socioeconomic drivers of chronic diseases, and guiding effective policy and intervention design aimed at promoting healthier aging populations worldwide.

Deep Learning Insights into Social Determinants of Chronic Disease and Longevity

Early, non-invasive detection of Alzheimer’s Disease (AD) significantly benefits public health planning, yet existing diagnostic tools are invasive, expensive, or clinic-dependent. This research develops a machine learning (ML) pipeline leveraging spontaneous speech, an accessible biomarker, using the balanced ADReSS 2020 dataset. We extracted 22 acoustic-prosodic features (including MFCCs, spectral centroid/bandwidth, RMS energy, fundamental frequency) combined with demographic metadata. After winsorizing outliers and applying standard scaling, four supervised ML algorithms (Gaussian Naïve Bayes, Logistic Regression, Support Vector Machine, and Gradient Boosting) were benchmarked. Optimal binning, novel to AD speech analysis, significantly enhanced Gradient Boosting performance, yielding 81.97% accuracy and ROC-AUC of 89.54%, outperforming other models by 7-30%. Feature analysis revealed pause- related spectral and pitch variability as key discriminators. Results indicate computationally efficient gradient boosting algorithms combined with optimal binning outperform complex deep learning models in interpretability and clinical feasibility, providing a strong foundation for practical, low-cost AD screening.

Downloads

Next from Aging Powerfully: AI Blueprint for Chronic Disease Management and Social Determinants of Longevity

AI-Based Early Osteoporosis Risk Detection Using Questionnaire Responses

Similar lecture

HeartLLM: Discretized ECG Tokenization for LLM-Based Diagnostic Reasoning

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES