Stroke and dementia are leading causes of morbidity and mortality worldwide, with disproportionately high burdens in sub-Saharan Africa (SSA). Despite this, early detection tools remain limited, particularly those adapted to Africa's cultural, environmental, and genomic diversity. We developed an AI-based diagnostic and risk assessment system to support early detection and risk stratification of stroke and dementia. The system integrates biomedical, cognitive, lifestyle, environmental, and psychosocial indicators. Cognitive assessment included Mini-Mental State Examination (MMSE) scores and an in-app memory recall game. Additional predictors included age, blood pressure, BMI, diabetes status, PTSD history, chronic pain, diet, sleep, education, physical activity, noise exposure, and stress scale scores. Seven machine learning algorithms were trained and validated. The dashboard was developed in Python using Streamlit framework and Supabase backend infrastructure, deployed online at https://huggingface.co/spaces/ademideola/african-neurohealth with multilingual support (Arabic, English, Swahili, French, Yoruba, Hausa, and Portuguese) to enhance accessibility across diverse African populations. For stroke diagnosis, the Random Forest and K-Nearest Neighbors models achieved the highest performance (Accuracy = 95.0%, AUC = 96–97%). For dementia prediction, Random Forest (Accuracy = 93.9%, AUC = 94.36%) and Decision Tree (Accuracy = 93.0%, AUC = 92.18%) outperformed other models. The developed dashboard demonstrates strong diagnostic and predictive performance for both stroke and dementia. By integrating cognitive, biomedical, and psychosocial parameters, this system offers a context-sensitive, AI-driven platform that supports early screening, individualized risk profiling, and public health decision-making in African populations.
Omolola et al. (Wed,) studied this question.