Equipping the GREDDIoMT Device with Early Behavioural Risk Detection of Dementia via a Pre-Activated SENet fused BiGRU

Gerontechnology has marked a transformative shift in healthcare via its fusion of embedded sensor units with Interne of Medical Things (IoMT), which embodies early disease detection and borders on its early warning for prevention. The IoMT equipped with sensors and algorithms, continuously monitors patients’ physiological data – provisioning real-time monitoring for early identification of potential health anomalies, and wades in with a personalized healthcare approach. With vast amount of data acquired, the gerontechs can be successfully trained with machine learning (ML) schemes, to gleans off insights via data analytics, and proffer medi-czars with proactive interventions for potential health risks prior its clinical manifestations. We posit a hybrid fusion of the dual channel squeeze-and-excite attention mechanism with bi-directional gated recurrent unit, for identifying the dementia disease. The model output is utilized as input for the GREDDIoMT (wearable) device that senses patient vitals, monitors and alerts on patients with dementia (PwDs), and yield resulting detection of the disease. With appropriate features selected for estimation, and imbalanced explored dataset resolved via SMOTEENN – our study shows that our hybrid (SENet+BiGRU) ensemble yields F1 0.995, Accuracy 0.997, Recall 0.998, Precision 1.000, AUC 0.997, and Specificity 1.000 – to accurately classify all 537-cases of test-dataset. In addition, our proposed hybrid model outperformed the various benchmarks. Keywords: Dementia, Healthcare, GREDDIoMT, SENet-BiGRU, Data Balancing, Disease detection, Journal Reference Format: Eferhire Ugbotu, Paul Onoma, Rita Ako, Anne Odoh, Duke Oghorodi, Ejaita Okpako, Tabitha Aghaunor, Frances Emordi, Emeke Ugboh, Joy Agboi, Chris Odiakaose, Arnold Ojugo, Victor Geteloma, Reuben Abere, Rebecca Idama, Andrew Eboka, Peace Ezzeh, Chris Onochie, Amanda Oweimieotu & Bright Ojo (2025): Equipping GREDDIoMT with Early Behavioural Risk Detection of Dementia via a Pre-Activated SENet fused BiGRU. Journal of Behavioural Informatics, Digital Humanities and Development Res. Vol. 11 No. 3. Pp 36-56. https://www.isteams.net/behavioralinformaticsjournal dx.doi.org/10.22624/AIMS/BHI/V11N3P4