Sleep is crucial for physical and mental recovery, and sleep positions significantly affect sleep quality. Most sleep trackers focus only on sleep patterns and stages, neglecting the role of sleep positions. To address this, we propose a wearable system with two wristbands and a chest-band that tracks sleep positions. It uses a two-level classifier within a Dynamic State Transition (DST) framework to analyze motion data from sensors and classify twelve motions across four sleep positions. Our tests show this system accurately tracks sleep positions, making it a vital tool for improving sleep care by focusing on sleep positions.

We developed a wearable system designed to detect strokes during sleep by analyzing asymmetrical motion with two wristbands. This system uses the EDIS (Early Detection of In-sleep Stroke) framework, leveraging deep learning and convolutional neural networks (CNNs) for efficient and accurate stroke detection. It processes wrist motion data into a format suitable for CNN analysis, leading to improved detection times and accuracy, notably with the EDIS-Resnet50 model demonstrating exceptional performance. This innovative approach aims to expedite stroke detection, potentially reducing the time to treatment and lowering the risk of severe outcomes.

Given the safety concerns and lack of proper screening for stroke survivors wishing to drive again, we propose Driving-PASS, a system utilizing driving simulators for safer and more informative assessment. This system aims to pre-screen potential stroke drivers and highlight specific areas for driving rehabilitation. It was developed using data from twenty-seven participants in thirteen scenarios, analyzed alongside assessments from ten experts across eleven driving factors. Driving-PASS employs eleven classifiers to evaluate driving abilities and suitability, overcoming data challenges with sophisticated feature extraction and data balancing techniques. This approach ensures a thorough and accurate assessment, promising to enhance road safety by effectively determining stroke survivors' fitness to drive.

In addressing the challenge of accurately predicting 28-day mortality for stage 4 cancer patients with septic shock, a machine learning (ML) model known as the balanced random forest (BRF) has been developed. This model significantly surpasses traditional scoring systems in predictive accuracy, as evidenced in a study with 897 patients where it achieved an area under the curve (AUC) of 0.821 in training and 0.859 in testing. The development of the BRF model marks a pivotal step in leveraging ML to enhance decision-making in the care of critically ill cancer patients, though further studies are needed for its global validation.

We've developed an IoT-based framework aimed at improving the treatment of patients experiencing shock, a condition demanding high medical expertise due to its complexity. Our system continuously monitors vital hemodynamic data, automatically detects shock states, and administers necessary medication, all while allowing medical professionals to track patient progress through a web application. This approach not only enhances patient care by ensuring timely interventions but also helps in minimizing the risk of transmitting infectious diseases among healthcare workers, offering a strategic solution to the current challenges in medical staffing and patient management in shock treatment

ROMI is a novel mobile robot developed for remote monitoring of intensive care unit patients during the COVID-19 pandemic, facilitating non-contact monitoring by healthcare professionals. It utilizes real-time optical digit recognition technology to interpret data from medical device LCD screens, such as those on ventilators and infusion pumps, thereby reducing the need for manual device checks. Created with Matlab Simulink and demonstrating high accuracy with the alexnet model, ROMI operates on NVIDIA Jetson hardware platforms. We evaluated the performance using 10 pre-trained CNN models and demonstrated accurate monitoring results that can aid in ICU patient care through an automated monitoring solution.

• Model-based and data-based approaches 

• Timeliness, Reliability, Safety, and Security

TBD