Canada

National Inter-University Health Data and AI Inquiry Program Abstract: Strokes are a leading cause of mortality in Canada and long-term disability worldwide, occurring either through blockage of blood or rupture of blood vessels in the brain. The impact of strokes is expected to increase as global populations age. Although clinical risk factors are prioritized in stroke prediction, lifestyle behaviours and social determinants of health (SDOH) also play a crucial role. Prevention of stroke not only affects populations on a personal and societal level but may also reduce its impact on healthcare systems. This project aims to aid individuals in predicting stroke risk by utilizing machine learning and providing preventative measures as well as promoting healthier aging based on public Canadian health data. The model uses Random Forest Regression to replace empty values in the dataset with an approximation. It tests the accuracy by predicting existing crude rate data and comparing the approximations to the actual value. The model demonstrated accurate predictions for lower crude rate values, but struggled with extreme values due to limited data availability. Results showed statistically significant relationships between stroke and various modifiable factors (p &lt; 0.05), including alcohol consumption, work stress, and physical activity. This project highlights the potential implementation of AI in primary stroke prevention using socioeconomic and behavioural data. In addition to clinical settings, this model can be applied on a larger scale in public health and insurance planning with further refinement and an expanded dataset.

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

Integrating Artificial Intelligence and Behavioural Risk Profiling for Early Stroke Prevention

socioeconomic

lifestyle

stroke

risk

machine learning

National Inter-University Health Data and AI Inquiry Program Abstract: Strokes are a leading cause of mortality in Canada and long-term disability worldwide, occurring either through blockage of blood or rupture of blood vessels in the brain. The impact of strokes is expected to increase as global populations age. Although clinical risk factors are prioritized in stroke prediction, lifestyle behaviours and social determinants of health (SDOH) also play a crucial role. Prevention of stroke not only affects populations on a personal and societal level but may also reduce its impact on healthcare systems. This project aims to aid individuals in predicting stroke risk by utilizing machine learning and providing preventative measures as well as promoting healthier aging based on public Canadian health data. The model uses Random Forest Regression to replace empty values in the dataset with an approximation. It tests the accuracy by predicting existing crude rate data and comparing the approximations to the actual value. The model demonstrated accurate predictions for lower crude rate values, but struggled with extreme values due to limited data availability. Results showed statistically significant relationships between stroke and various modifiable factors (p < 0.05), including alcohol consumption, work stress, and physical activity. This project highlights the potential implementation of AI in primary stroke prevention using socioeconomic and behavioural data. In addition to clinical settings, this model can be applied on a larger scale in public health and insurance planning with further refinement and an expanded dataset.

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. 

Accurately detecting Alzheimer’s disease (AD) in its early stages is critical for intervention yet remains a persistent clinical challenge. Although volumetric MRI biomarkers like hippocampal atrophy are well-established, emerging research suggests that radiomic texture features may offer insights into neurodegeneration. This study investigates the relative predictive value of volumetric, demographic, and gray-level co-occurrence matrix (GLCM)-based texture features for early AD classification using the OASIS-2 dataset. T1-weighted MRI scans and clinical metadata from 120 participants (48 demented, 72 nondemented, excluding converted) were analyzed. Radiomic features, including GLCM texture metrics and first-order statistics, were extracted from skull-stripped sagittal slices. Volumetric estimates were derived from brain masks, and demographic variables included age, sex, education, and socioeconomic status (SES). Multiple Random Forest classifiers were trained using combinations of feature groups. Hyperparameter optimization was performed using grid search with 5-fold cross-validation, and class imbalances were addressed by the utilization of Synthetic Minority Oversampling Technique (SMOTE). The best performing model was the combination of volumetric and demographic features, achieving an accuracy of 79.2% and an AUC of 0.8430. GLCM texture features alone yielded lower predictive performance (accuracy: 57.5%, AUC: 0.5784) but provided marginal benefit when added to volume and demographic data (accuracy: 73.33%, AUC: 0.7935). Confusion matrix and Receiver Operating Characteristic (ROC) analyses confirmed the superiority of multimodal models. However, texture features may have been limited by slice-based extraction and modest sample size. These findings support the clinical utility of integrating structural MRI with demographic profiling for scalable, non-invasive AD screening.

Machine Learning Driven Discovery of MRI-based Alzheimer’s Biomarkers

DNA methylation is an epigenetic modification that involves the addition of methyl groups to DNA at specific sites, often affecting gene expression. These patterns can be influenced by genetic and environmental factors, including aging. Specific DNA methylation sites have been linked to various diseases, such as Alzheimer’s disease and cancer, making them promising epigenetic biomarkers for age and disease prediction. This study aims to develop machine learning models to predict chronological age and classify Alzheimer’s disease using DNA methylation profiles derived from blood samples. Two datasets were made from publicly available studies: the age prediction dataset included 9 studies with a total of 4,303 samples (median age 54 years) and the Alzheimer’s disease dataset which included 2 studies comprising of 823 samples, 247 of whom were from individuals diagnosed with Alzheimer’s. The final age prediction model achieved a mean absolute error (MAE) of 3.01 years and a median absolute error (MedAE) of 2.40 years. The Alzheimer’s classification model attained an accuracy of 79%, with a false positive rate of 13.9% and a false negative rate of 36.6%. Both models identified sites associated with relevant genes. The age model highlighted ELOVL2, FHL2, KLF14, and TRIM59, which are commonly associated with aging, while the Alzheimer’s model identified SCGN, FKBP5, GATA4, ORMDL3, IRF3, and SPIDR, all of which have been linked to Alzheimer’s disease in previous studies. These findings support the potential of DNA methylation as an epigenetic biomarker for both age and disease prediction.

Machine Learning Prediction of Age and Alzheimer’s Disease from DNA Methylation Profiles

Alzheimer’s disease (AD) is a prominent form of dementia, accounting for 60–70% of all dementia cases. AD has many different risk factors, with the most prominent one being age - meaning it is a major concern for the aging population. In recent years, chatbots have become a popular and interactive form of software with the capability to perform certain actions. In the context of AD, chatbots can be leveraged to provide an accessible interface for machine learning classification models. This paper explores using Rasa, an open-source framework with natural language processing capabilities, as an interactive platform that can switch between machine learning models trained to classify different types of data based on identified intent. Two separate datasets were used to represent distinct diagnostic modalities: one with clinical and demographic risk factor data, and another containing magnetic resonance imaging (MRI) brain images. A Random Forest Classifier achieved 95% accuracy and 0.96 area under the curve (AUC) on the tabular dataset, while a Convolutional Neural Network reached 96% accuracy on the MRI dataset, though performance varied by class, with lower AUC for the “Very_Mild_Dementia” class. The proposed chatbot interface facilitates intuitive use, supports model extensibility, and reduces the need for extensive software training among clinicians. This approach demonstrates that a chatbot-interfaced application can effectively reduce the technological knowledge required to leverage machine learning for Alzheimer's disease diagnosis.

Multimodal Machine Learning-Based Diagnosis Chatbot for Alzheimer’s Disease

Cognitive decline and dementia are increasing public health concerns among Canada’s aging population. Musical interaction has been a potential non-pharmacologic intervention to enhance cognitive and affective well-being. Its benefits may not be widely distributed to all groups, though. This research investigates the association between musical engagement and cognitive function, and social determinants among older Canadians using analysis of two national surveys: the Canadian Community Health Survey (CCHS – Healthy Aging) and the General Social Survey (GSS – Time Use, Cycle 29). Affordable artificial intelligence (AI) techniques, including multiple linear regression and k-means clustering, were employed to explore predictors of depression and musical engagement trends. Depression scores on the Patient Health Questionnaire-9 (PHQ-9) were approximated based on loneliness, income, education, race, and immigration status. Clustering pinpointed high depression groups with limited music access. GSS data illustrated differences in gender and income participation in music despite regression findings indicating poor predictive capability (R² ≈ 0.00). Loneliness was the strongest predictor of depression (coef = 0.36), and the model explained 34% of the variance (R² = 0.34). Findings indicate that racialized, low-income older adults are least likely to have music-based care. These findings identify that AI can uncover disparities and guide culturally responsive, community-based programming to promote equity in healthy aging.

Musical Engagement and Cognitive Risk: an Equity-Focused Analysis in Canadian Seniors

Across North America, elders face high rates of hypertension, type 2 diabetes, and hypercholesterolemia, yet standard DASH guidelines lack the precision to accommodate age, sex, and comorbidity differences. We developed a linear‐programming engine that transforms DASH nutrient ranges into individualized meal plans for 18 geriatric cohorts (female/male × three age bands × three health statuses). Using an 8 × 49 food‐group incidence matrix and cohort‐specific lower/upper bounds drawn from NHLBI DASH targets, energy‐scaling factors, and disease caps, our model minimizes total slack across all constraints. Implemented in Python 3.11 with PuLP and the HiGHS dual‐simplex solver, it produced feasible menus for every cohort, with maximum deviation ≤ 0.3 servings per group. Key findings include robust compliance with fruit and vegetable targets, while grains and fats/oils bounds were most frequently binding—especially in oldest and hypercholesterolemic cohorts. A ±10 \% sensitivity test showed that modest bound adjustments markedly reduce average slack yet shift which groups become most challenging. Our approach demonstrates that precision nutrition via linear optimization can operationalize DASH into concrete, cohort‐tailored menus, offering a transparent, reproducible framework for personalized geriatric dietary planning.

Optimizing Geriatric DASH Diet with Linear Programming for Personalized Nutrition

Rheumatoid Arthritis (RA) is the most common chronic autoimmune joint disease and a major cause of disability globally, particularly impacting the elderly with premature immune aging and a reduced life expectancy of up to 10 years. Early diagnosis is critical to improving outcomes, yet current diagnostic methods rely on time-consuming and costly manual scoring of joint damage from X-ray images using the Sharp/van der Heijde (SvH) system, an indicator of disease severity. In this study, we present a deep learning pipeline utilizing the EfficientNet-b0 convolutional neural network (CNN) to automate RA classification from hand X-ray images, using binary classification of patients as either “Healthy” or “RA” based on a predefined Sharp/van der Heijde (SvH) score threshold. Our approach integrates a robust preprocessing pipeline (grayscale conversion, Gaussian denoising, CLAHE, and specialized normalization) and a two-phase transfer learning strategy for model training. On the testing dataset, the model achieved a test accuracy of 83.77%, RA precision of 80.00%, RA recall of 53.85%, AUC-ROC of 88.54%, specificity for the healthy class of 94.96%, and an RA F1-score of 64.37%. These results highlight the model’s potential to aid early RA diagnosis and serve as a foundation for AI-based diagnostic tools in aging populations, supporting timely intervention and improved longevity.

Optimizing Radiographic Scoring in Elderly Patients with Rheumatoid Arthritis Using a Convolutional Neural Network Deep Learning Model

The CatchIt Early research project examined how artificial intelligence (AI) and machine learning (ML) could support early detection of neurodegenerative disorders such as Alzheimer’s and Parkinson’s disease in Canada. Data from the Canadian Community Health Survey (CCHS) survey was used to model early indicators of cognitive decline in diagnosis based on five (5) variables: perceived mental health, diagnosed mood disorder, sleep issues, mobility limitations, and memory issues. Following preprocessing and dimensionality reduction, random forest and logistic regression learning models were trained to classify individuals at risk. Both models performed poorly, with area under the curve (AUC) values of 0.507 and 0.505, low specificity, and negligible Cohen’s Kappa scores. While sensitivity was moderate, both models failed to distinguish true positives from false positives reliably. These findings indicate that self- reported data alone is insufficient for early detection of neurodegeneration using current AI/ML approaches. To contextualize the prevalence, the choropleth maps and a line graph were created to illustrate a consistent rise in Alzheimer’s cases from 2002 to 2022 and Parkinson’s from 2006 to 2022, with notable provincial variation. While predictive modeling using existing public health data was unsuccessful, the project emphasizes the urgent need for richer, longitudinal, and clinically validated datasets to improve early detection capabilities. Ultimately, the study highlights the growing burden of neurodegenerative disorders in Canada and the importance of advancing data integration to support proactive public health planning and early intervention.

Predicting Alzheimer’s and Parkinson’s Onset Through Machine Learning and Public Health Data

As populations age, early detection of chronic illness is increasingly vital to prevent declines in independence and quality of life. This study investigates whether wearable-derived physiological signals and clinical health records can be used to predict chronic illness in older adults using machine learning models. We used data from the All of Us (AOU) Research Program, including Fitbit wearable summaries and EHR-based clinical measures, to predict three high-burden chronic conditions: osteoporosis, dementia, and heart failure. After filtering for data completeness, we created participant-level summaries using 21 features spanning physical activity, sleep behavior, cardiovascular metrics, and biometric indicators. A final dataset of approximately 2,000 participants aged 60+ was used for modeling. Logistic regression, random forest, and XGBoost models were trained and evaluated using 5-fold cross-validation with stratified train/test splits. We addressed class imbalance with up-sampling during training. Performance was measured using accuracy, F1-score, and area under the ROC curve (AUC). Osteoporosis prediction achieved the highest performance across models (XGBoost AUC = 0.802, F1 = 0.742), supported by features such as basal metabolic rate (BMR) and daily energy expenditure. Dementia and heart failure results were limited by severe class imbalance, with undefined F1-scores and reduced AUCs. These results demonstrate the feasibility of predicting musculoskeletal chronic conditions using wearable and clinical data, and highlight the challenges of modeling rare disease outcomes in imbalanced datasets. Our findings suggest a scalable framework for early risk stratification using passive health monitoring in aging populations.

Predicting Chronic Illness in Aging Populations Using Physiological Markers

Predicting Loneliness-Linked Health Risks by Analyzing Digital Behaviours Using AI

As public health and demographic shifts coexist, aging populations present challenges for disease prediction and prevention in developing regions. Focusing on the urban and rural areas of Punjab, Pakistan, and Kandahar and Kabul, Afghanistan, the presented research approaches the task of developing a predictive AI framework to quantify polio and hypertension risk. Bayesian networks were employed to model the probability of causal relationships between the parameters of age, vaccination status, sanitation, urbanity or rurality, and water access. Synthetic, yet realistic, data was integrated to reflect true conditions necessary for training the model to understand real data that was later provided. Risk calculations produced capture the multiplicative effects of low vaccination statuses, poor sanitation, and aging on the probability of being at risk. Results show regional differences due to urbanity and rurality showing higher risk of hypertension in the latter, but a main focus on poor sanitation correlating to higher risk of polio. Hypertension is shown to have a positive correlational relationship with aging, while polio is shown to have a negative correlational relationship. The AI model thus shows how localized interventions like, sanitation campaigns or an increase in elderly care, could mitigate risks. These findings underline a need for data-driven public health strategies in resource-scarce environments. The framework’s ability to adapt to parameters suggests applications in other regions, age groups, diseases, and more, provided that real data is used. This contributes to a replicable method for health interventions where inequities reside.

Downloads

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

Machine Learning Driven Discovery of MRI-based Alzheimer’s Biomarkers

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES