212 research outputs found
Dynamic and Explainable Mortality Risk Prediction for TBI Patients in the ICU
No full textDynamic mortality risk prediction in the intensive care unit (ICU) is crucial for supporting clinicians’ decision-making, specifically in traumatic brain injury (TBI) patients. We aim to develop and evaluate a dynamic deep learning (DL) framework that can provide hourly updates of 30-day mortality risk prediction for TBI patients following ICU admission. Using demographics and time-series physiological data, a recurrent neural network-based model was trained on data from 135 TBI patients admitted to the Gold Coast University Hospital (GCUH) in Australia. Model’s performance was evaluated utilizing the area under the receiver operating characteristics (AUC), Matthews correlation coefficient (MCC), accuracy, and other metrics, performed calibration and decision curve analysis to interpret the model’s output and determine its clinical usefulness. The Shapley additive explanation algorithm was utilized to clarify the contribution of features to the predictions. The proposed method showed predictive performance on the cross-validation dataset that improved over time: MCC 0.24 and AUC 0.713 for the prediction at 24 h after admission, 0.451 and 0.756 at 72 h, 0.519 and 0.803 at 120 h, and 0.748 and 0.946 before twelve hours to the outcome (either death or discharge), respectively. The model was further tested with a holdout test dataset with 34 TBI patients, achieving an average prediction accuracy of 0.851, AUC of 0.632, and MCC of 0.403, respectively, in the first 24-h interval. The proposed model demonstrates proof of principle with explainable results in predicting mortality risk, encouraging further development and validation in a clinical setting
Spatiotemporal Contrastive Learning for Echocardiography View Classification
No full textEchocardiographic view classification is essential for accurate cardiac assessments, yet it remains challenging due to anatomical overlap, operator variability, motion artifacts, image quality issues, and dataset limitations. Deep learning methods could address these issues by incorporating temporal models, representation learning, and domain adaptation to improve classification robustness. This study proposes a contrastive representation learning framework that integrates temporal and spatial augmentation strategies, to learn more robust and invariant feature representations. Experimental results demonstrate that the proposed approach achieves an accuracy of 96.4%, surpassing previous methods. The findings indicate that the model effectively captures robust and invariant feature representations, strengthening its ability to distinguish between echocardiographic views and consequently enhancing classification performance
MICA : a multimodal intelligent cognitive assessment framework integrating generative AI and social robot for early cognitive intervention
No full textThe rapidly growing older adult population underscores the urgent need for innovative solutions to detect, classify, and monitor early cognitive decline. Traditional cognitive screening methods, such as paper-and-pencil tests like the Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination (MMSE), suffer from notable limitations, including low patient engagement, limited motivational appeal, inadequate sensitivity to subtle cognitive decline, and susceptibility to practice effects with repeated administrations. Furthermore, such tests require substantial clinician time, as they must be administered by trained healthcare practitioners, increasing healthcare costs. Given the current shortage of effective interventions and reliable screening methods for early-stage cognitive decline, socially assistive robots offer a promising dual function: They can deliver personalized and engaging cognitive stimulation and social support while monitoring cognitive health. This paper proposes a new Multimodal Intelligent Cognitive Assessment (MICA) framework integrated into the Pepper social robot and enhanced by generative AI technologies. MICA consists of three core components:(1) a conversational and cognitive exercise interface powered by generative AI, enabling natural, engaging interactions tailored to various cognitive domains; (2) multimodal perception capabilities, incorporating emotion recognition using DeepFace, robust speech recognition, and real-time personalized adaptation based on emotional and cognitive feedback, and (3) an advanced performance logging system designed to systematically record patient accuracy, response time, and emotional states. Initial evaluations with Pepper demonstrated real-time emotion detection and adaptive exercises, which illustrate the potential for high levels of engagement and early intervention. Although current evaluations were conducted by the authors, more comprehensive user studies are planned to validate the effectiveness of MICA within the populations of interest
- …
