Chao-Tung Yang

Protein language models have achieved remarkable success in structure prediction and de novo design. However, their understanding of physical interactions remains limited. For structure-based models, we hypothesize that this limitation stems from the absence of explicit solvent modeling: water mediates hydrogen bonding, hydrophobic burial, and electrostatic interactions that govern protein thermodynamics and kinetics, yet current structure-based models treat proteins as if they exist in a vacuum. Using thermophilic protein classification as a proxy task for force-field awareness, per-residue logit mis-correlated with H-bonds as a destabilizing attribute and exhibits significant solvent-accessible surface area (SASA)-dependent error, despite high overall classification accuracy. These artifacts point to the water–protein interface as a source of systematic bias. To address this, we propose HEIWA (HybridExplicit-Implicit Water Attention), a novel attention module that injects solvent-mediated force-field information into static ESM3 residue embeddings via an all-query-to-all-key water–residue attention. Biophysics-guided model selection identifies configurations whose residue-level outputs are more consistent with the RMSF benchmark and selected biophysical attributes, while classification performance remains broadly comparable to the baseline. By forcing the model to predict the protein thermophilicity from the contributions of individual residues, HEIWA learns biophysical patterns that can be extended to SUPFAM absent from the training data set. HEIWA also largely preserves the embedding structure of ESM3. Molecular dynamics simulations show improved kinetic stability under thermophilic conditions, suggesting the potential to generate structures with greater kinetic stability. To our knowledge, this is the first study to incorporate explicit solvent encoding into a residue-resolution protein language model.

Background: Tibial plateau fractures are complex knee injuries where timely and accurate diagnosis is critical to preventing long-term disability. In high-pressure emergency settings, the risk of missed fractures (false negatives) remains a significant challenge. Objective: This study aims to develop a robust, clinically safe automated detection model using advanced deep learning architectures. Methods: We utilized a dataset of 1489 real-world clinical X-ray images, annotated by orthopedic surgeons, to train and evaluate five versions of the You Only Look Once (YOLO) algorithm (v8, v9, v10, v11, and v12). A rigorous two-stage evaluation process was implemented. First, an initial screening excluded YOLOv8 and YOLOv10 due to critical detection failures (“background errors”), in which the models failed to detect any object in the target region. Second, a comprehensive performance analysis identified YOLOv11 as the optimal architecture. Based on these results, the YOLOv11 model was integrated into a user-friendly, web-based diagnostic system using the Python Flask framework. Results: The YOLOv11 model achieved the highest Mean Average Precision (mAP) of 99.3% and an Accuracy of 98.32%. Crucially for clinical safety, YOLOv11 demonstrated superior sensitivity (96.58%) with the lowest false negative rate, attributed to its enhanced feature aggregation capabilities which effectively distinguish subtle fracture lines from trabecular bone patterns. Independent web interface validation (n = 109 real-world cases) confirmed 98.17% accuracy, 98.00% sensitivity, 98.31% specificity, 98.00% Positive Predictive Value (PPV), and 98.31% Negative Predictive Value (NPV). Conclusions: This system is designed to support clinical workflows by providing real-time, highly accurate second opinions, thereby reducing diagnostic errors and alleviating radiologist workload.

The default Horizontal Pod Autoscaler (HPA) in Kubernetes is inherently reactive, often taking 15–30 seconds to respond to workload changes on resource-constrained edge devices. In latency-sensitive applications such as real-time crowd analytics, these delays frequently cause thermal throttling and QoS violations during sudden surges in crowd density. To overcome this limitation, we propose Edge-VisForeCast, a proactive autoscaling framework that leverages YOLOv11-detected crowd density as a Granger-causal leading indicator of future resource demand. By integrating these application-level insights with a quantized Informer transformer model enhanced by a Vision-Aware Attention mechanism, the system can forecast power consumption 5–15 seconds ahead. Granger causality tests (p < 0.001) confirm that person count statistically precedes CPU and power saturation by 10–15 seconds. We deployed the proposed system on a 3-node NVIDIA Jetson Orin Nano cluster at the Tunghai University Library, using a real-world dataset of 39.01 million raw frames collected over 14 days. Experimental results show that Edge-VisForeCast achieves 17.8% better forecasting accuracy (MAE) than LSTM baselines, while delivering a 23.7% reduction in inference latency and a 36.3% reduction in energy consumption. Most importantly, the system maintains stable operation within the 15W thermal envelope, effectively avoiding the throttling issues commonly observed with reactive HPA. This work demonstrates the value of bridging application semantics with system orchestration for energy-efficient predictive autoscaling in IoT edge environments.

Objectives: Ensuring an effective triage to treat patients with chest pain in emergency settings is critical, but it can often be challenging, particularly when patients wear face masks or are unable to clearly communicate their pain. To address this limitation, this study presents a real-time facial expression–based system for chest pain intensity assessment as an initial step toward realizing intelligent emergency triage. The proposed system integrates deep learning with real-time video analysis to provide objective and rapid pain level recognition. Methods: A YOLOv12-based facial expression recognition model was trained using annotated facial images of patients experiencing chest pain, and the model categorizes pain into three intensity levels: no pain, slight pain, and moderate to severe pain. Multiple YOLOv12 variants were systematically evaluated to identify an optimal configuration for potential clinical use. The developed system supports two operational modes: real-time recognition, which analyzes continuous video streams and delivers immediate visual feedback through an interactive interface, and a manual upload mode for offline video analysis, review of results, and playback. Additional usability features, including error prompts and data reset functions, were implemented to enhance system stability and user experience. Results: Among the evaluated models, the YOLOv12-L model achieved the best performance with an accuracy of 98.81%, sensitivity of 98.76%, specificity of 98.79%, precision of 98.04%, and an F1-score of 98.41%, demonstrating stable and accurate recognition. The proposed system is designed to support the triage process of assessing patients with chest pain, particularly in cases where patients wear masks or cannot clearly express their pain. By providing real-time and objective pain intensity assessment, the system shows potential to assist healthcare professionals in identifying patients who may require priority attention and to serve as a supportive tool for emergency triage workflows. Conclusions: Future work will incorporate edge computing with a lightweight model to enable real-time pain assessment in ambulances, facilitating faster intervention and treatment.