Virtual reality (VR) Rehabilitation holds the potential to address the challenge that patients feel bored and give up long-term rehabilitation training. Despite the introduction of gaming elements by some researchers in rehabilitation training to enhance engagement, there remains a notable lack of in-depth research on VR rehabilitation serious game design methods, particularly the absence of a concrete design framework for VR rehabilitation serious games. Hence, we introduce the Clinical-Function-Interesting (CFI): a VR rehabilitation serious game design framework, harmonizing rehabilitation function and game design theories. The framework initiates with clinic information, defining game functions through the functional decomposition of rehabilitation training. Subsequently, it integrates gaming elements identified through the analysis and comparison of related literature to provide enduring support for long-term training. Furthermore, VR side-effect and enhancement are considered. Building upon this design framework, we have developed an upper limb VR rehabilitation serious game tailored for mild to moderate stroke patients and aligned our framework with another developed VR rehabilitation serious game to validate its practical feasibility. Overall, the proposed design framework offers a systematic VR rehabilitation serious game design methodology for the VR rehabilitation field, assisting developers in more accurately designing VR rehabilitation serious games that are tailored to specific rehabilitation goals.

With the improvement of the quality of life, the demands on button mushroom (Agaricus Bisporus) increase significantly. But traditional farming methods mainly rely on costly human picking, which is currently facing the time-consuming and labor-intensive problem. As machine vision technology becomes more mature, harvesting robot with flexible picking end-effector is an alternative method to address this issue. Though the conventional harvesting robot is capable of mushrooms picking, but the damage rate is relatively high for the improper picking force and motion. Thus, a novel picking end-effector for button mushrooms is designed based on vacuum negative pressure picking in this paper. The harvesting robot with flexible end-effector is proposed to solve the problem: 1) to avoid the injury when the end-effector touch the mushroom; 2) to increase the picking efficiency without damage rate increase. The structure of robotic mushroom picking end-effector is described in detail, and then the kinematic modeling and picking force analysis of this robot are presented. Lastly, the numerical simulation via the MATLAB is carried out to study the influence of the robot parameters. The bruise tests indicated that the maximal allowable stress is 0.196 MPa and mean allowable pressure is 13.82 N. Preliminary results demonstrate that the robot achieved 88.2% success rate and 2.9% damage rate in the factory environment, and there is potential for the automatic mushroom harvesting with the proposed harvesting robot.

A cornerstone of the rehabilitative regime for people diagnosed with dyskinesia is walking assist and gait training, however, prolonged care, provided by relatives or professionals, serves as a massive financial burden to patients. The proposed novel robotic walker seeks to address this concern. The walker incorporates human motion intention recognition to facilitate lower limb rehabilitation training and daily walking. The walker design places strong emphasis on patient safety and quality of life by providing omni-directional walking assist, four key pelvic motions that support hip rotations and comfortable body weight support (BWS). Additionally, five sensors were installed to identify the user's motion intention from the interaction forces surrounding the pelvis and dead zone. Furthermore, Kalman filtering was used to guarantee the quality of the interactive signal while kinematic and dynamic models were derived to generate appropriate driving velocities to support patient's body weight and improve mobility. To validate our design, the MATLAB simulations and exploratory clinical trials using healthy subjects were performed. Preliminary results demonstrate satisfactory kinematic performance and suggest the walker as a promising therapeutic avenue for individuals suffering from dyskinesia and other associated movement disorders.

In response to the ever-increasing demand of community-based rehabilitation, a novel smart rehab walker iReGo is designed to facilitate the lower limb rehabilitation training based on motion intention recognition. The proposed walker provides a number of passive degrees of freedom (DoF) to the pelvis that are used to smooth the hip rotations in such a way that the natural gait is not significantly affected, meanwhile, three actuated DoF that are actively controlled to assist patients with mobility disabilities. The walker first identifies the user's motion intention from the interaction forces in both left and right sides of the pelvis and then uses the kinematic model to generate appropriate driving velocities to support the body weight and improve mobility. In this paper, workspace, dexterity and the force field of the walker are analyzed based on the system Jacobian. Simulation and experiments with healthy subjects are carried out to verify the effectiveness and tip-over stability. These results demonstrate that the walker has sufficient workspace for pelvic motions, satisfactory dexterity and near-linear force feedback within the prescribed workspace, and that the walker is easily controlled to ensure normal gait.

Facial emotion recognition has become increasingly important in healthcare, where understanding delicate cues like pain, discomfort, or unconsciousness can support more timely and responsive care. Yet, recognizing facial expressions in real-world settings remains challenging due to varying lighting, facial occlusions, and hardware limitations in clinical environments. To address this, we propose TriViT-Lite, a lightweight yet powerful model that blends three complementary components: MobileNet, for capturing fine-grained local features efficiently; Vision Transformers (ViT), for modeling global facial patterns; and handcrafted texture descriptors, such as Local Binary Patterns (LBP) and Histograms of Oriented Gradients (HOG), for added robustness. These multi-scale features are brought together through a texture-aware cross-attention fusion mechanism that helps the model focus on the most relevant facial regions dynamically. TriViT-Lite is evaluated on both benchmark datasets (FER2013, AffectNet) and a custom healthcare-oriented dataset covering seven critical emotional states, including pain and unconsciousness. It achieves a competitive accuracy of 91.8% on FER2013 and of 87.5% on the custom dataset while maintaining real-time performance (~15 FPS) on resource-constrained edge devices. Our results show that TriViT-Lite offers a practical and accurate solution for real-time emotion recognition, particularly in healthcare settings. It strikes a balance between performance, interpretability, and efficiency, making it a strong candidate for machine-learning-driven pattern recognition in patient-monitoring applications.

The increasing adoption of artificial intelligence (AI) in intelligent healthcare systems has elevated the demand for robust medical imaging and vision-based inventory solutions. For an intelligent healthcare inventory system, accurate recognition and classification of medical items, including medicines and emergency supplies, are crucial for ensuring inventory integrity and timely access to life-saving resources. This study presents a hybrid deep learning framework, EfficientDet-BiFormer-ResNet, that integrates three specialized components: EfficientDet’s Bidirectional Feature Pyramid Network (BiFPN) for scalable multi-scale object detection, BiFormer’s bi-level routing attention for context-aware spatial refinement, and ResNet-18 enhanced with triplet loss and Online Hard Negative Mining (OHNM) for fine-grained classification. The model was trained and validated on a custom healthcare inventory dataset comprising over 5000 images collected under diverse lighting, occlusion, and arrangement conditions. Quantitative evaluations demonstrated that the proposed system achieved a mean average precision (mAP@0.5:0.95) of 83.2% and a top-1 classification accuracy of 94.7%, outperforming conventional models such as YOLO, SSD, and Mask R-CNN. The framework excelled in recognizing visually similar, occluded, and small-scale medical items. This work advances real-time medical item detection in healthcare by providing an AI-enabled, clinically relevant vision system for medical inventory management.