Posture Monitoring and Classification for Health Care of Bedridden Patients Using IoT and AI Algorithms

Posture monitoring has become a crucial area of research and application, particularly considering the increasing concerns about musculoskeletal disorders and their association with prolonged lying and poor ergonomic practices. To solve the crucial problem of bedridden patient care, this study investigates the categorization of lying postures using four different deep learning models: The models used are Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), VGG16, and ResNe$t$50. Due to bedridden patients' difficulty in developing pressure ulcers, respiratory disorders, and other complications, the classification of laying positions which includes supine, prone, and lateral is essential. Some possible assistive devices that can inform caregivers or initiate some techniques to improve the patient's comfort and eliminate adverse effects of immobility can benefit from the identification of some postures in real-time. The models were trained using both balanced and unbalanced data sets from a set of labeled images covering different laying postures. Their performance was then determined by criteria including accuracy, RMSE, Kappa, recall, and precision. As for the performance, the proposed VGG model achieved better classification accuracy compared to the CNN, RNN, and ResNet models. In light of enhancing patient care as well as monitoring within the health care facilities, it is suggested from our result that these models could suggest potential solutions to posture identification for bedridden patients. By illustrating how AI posture classification systems can be implemented in IoT-based applications, the study also shows how future healthcare delivery can be improved and the problems caused by long periods of inactivity can be prevented.