Background/Objective: Muscle synergy analysis based on machine learning has significantly advanced our understanding of the mechanisms underlying the central nervous system motor control of gait and has identified abnormal gait synergies in stroke patients through various analytical approaches. However, discrepancies in experimental conditions and computational methods have limited the clinical application of these findings. This review seeks to integrate the results of existing studies on the features of muscle synergies in stroke-related gait abnormalities and provide clinical and research insights into gait rehabilitation. Methods: A systematic search of Web of Science, PubMed, and Scopus was conducted, yielding 10 full-text articles for inclusion. Results: By comprehensively reviewing the consistencies and differences in the study outcomes, we emphasize the need to segment the gait cycle into specific phases (e.g., weight acceptance, push-off, foot clearance, and leg deceleration) during the treatment process of gait rehabilitation and to develop rehabilitation protocols aimed at restoring normal synergy patterns in each gait phase and fractionating reduced synergies. Conclusions: Future research should focus on validating these protocols to improve clinical outcomes and introducing indicators to assess abnormalities in the temporal features of muscle synergies.

Background: Predicting a stroke in advance or through early detection of subtle prodromal symptoms is crucial for determining the prognosis of the remaining life. Electromyography (EMG) has the advantage of easy and quick collection of biological data in clinical settings; however, its application in data processing and utilization is somewhat limited. Thus, this study aims to verify how simple signal processing and feature extraction utilize EMG in machine learning (ML)-based prediction models. Methods: EMG data were collected from the legs of 120 healthy individuals and 120 stroke patients during gait. Four statistical features were extracted from 16 EMG signals and trained on seven ML-based models. The accuracy of the validation and test datasets was also examined. Results: The model with the best performance was Random Forest. Among the 16 EMG signals, the average and maximum values of the muscle activities involved in knee extension (i.e., vastus medialis and rectus femoris) contributed significantly to the predictions. Conclusion: The results of this study confirmed that the simple processing and feature extraction of EMG signals effectively contributed to the accuracy of ML-based models. Routine use of EMG data collected in clinical environments is expected to provide benefits in terms of stroke prevention and rehabilitation.

Objective: Ulcerative colitis (UC) endoscopic image classification presents challenges owing to imbalanced medical imaging data, particularly when the clinical importance of accurate positive predictions increases with disease severity. This study proposes a multi-task learning (MTL) framework inspired by the coarse-to-fine processing mechanism of the human brain to address these challenges. Methods: The proposed MTL framework was evaluated using endoscopic images of UC, focusing on its ability to classify disease stages with an emphasis on the accurate detection of advanced cases. Results: Our findings demonstrate that the proposed framework effectively mitigates the limitations posed by imbalanced datasets, particularly by enhancing classification performance in severe disease stages. Notably, DenseNet121 exhibited significantly superior performance compared to other backbones and achieved an additional performance gain in identifying Mayo endoscopic scores of 2 and 3 following joint-loss optimization. Additionally, MobileNet-v3-large, despite being a lightweight model, demonstrated notable gains under the proposed optimization scheme, highlighting the versatility of the framework across architectures with different computational complexities. Conclusion: MTL-based computer-aided diagnosis enables conservative and accurate identification of patients in critical stages, supporting timely and appropriate treatment decisions while reducing the risk of underdiagnosis and delayed care. Furthermore, our results highlight the potential of MTL in overcoming data imbalance issues. Future studies should explore integrating multiple convolutional neural network-based models to further boost classification accuracy.

Colorectal cancer (CRC) has a relatively high five-year survival rate compared to other cancers; however, this rate drops significantly in patients with malignant CRC. One critical factor in palliative care decision-making is the ability to accurately predict patient survival, with the six-month survival period commonly used as a threshold. In this study, we evaluated the performance of five machine learning models—logistic regression, decision tree, random forest, multilayer perceptron, and extreme gradient boosting (XGBoost)—in predicting six-month survival for patients with malignant CRC using a publicly available synthetic dataset containing 11,774 samples and 51 features. The models were trained and validated using five-fold cross-validation, and the synthetic minority oversampling technique (SMOTE) was applied to address class imbalance. Among the models, XGBoost demonstrated the highest performance, achieving 95% accuracy, precision, recall, and F1-score, along with 90% specificity. Feature importance analysis identified smoking status and surgical history as key factors influencing model predictions. These findings highlight the potential of tree-based machine learning models in supporting timely and informed palliative care decisions, while also providing insights into handling data imbalance and optimizing model parameters in survival prediction tasks.

BACKGROUND Wearable photoplethysmography (PPG) enables continuous ambulatory physiological monitoring; however, heart rate variability (HRV) estimation under free-living conditions remains challenging because of motion artifacts, signal instability, and heterogeneous activity-dependent physiological variability. In particular, robust subject-independent wearable HRV estimation without user-specific calibration remains insufficiently explored. OBJECTIVE This study investigated a context-aware wearable HRV estimation framework integrating signal-quality-aware and motion-aware contextual information for subject-independent ultra-short-term HRV estimation across heterogeneous daily activities. METHODS Wrist PPG and accelerometer signals from the publicly available PPG-DaLiA dataset were analyzed under subject-independent evaluation conditions. The proposed framework integrated PPG waveform representations with two forms of contextual information: (1) signal-quality-aware contextual features representing wearable signal reliability and (2) accelerometer-derived motion-aware contextual information representing ambulatory movement intensity. Unlike approaches relying on explicit activity labels, the proposed framework used lightweight contextual representations derived directly from wearable sensor signals. Four experimental configurations were compared: PPG-only, PPG with signal-quality-aware contextual information, PPG with motion-aware contextual information, and PPG with full context-aware information. SDNN was used as the primary HRV target. RESULTS Signal-quality and motion-related cues influenced wearable HRV estimation performance under subject-independent free-living conditions. The signal-quality-aware contextual configuration achieved the strongest overall estimation performance (Pearson r=0.548, MAE=0.0142 s, RMSE=0.0194 s), outperforming the PPG-only baseline configuration (Pearson r=0.329, MAE=0.0199 s, RMSE=0.0245 s). Motion-aware contextual information also improved overall trend tracking performance relative to the baseline model (Pearson r=0.444). However, contextual effects varied considerably across activities. Cycling and lunch break activities demonstrated relatively improved contextual estimation performance, whereas stair-related activities remained challenging across all configurations, with near-zero or negative correlation coefficients despite modest absolute estimation errors. Visualization and agreement analyses further demonstrated substantial differences in prediction behavior and estimation stability across contextual configurations. CONCLUSIONS Contextual information substantially affected wearable HRV estimation robustness under heterogeneous free-living conditions. Signal-quality-aware contextual information demonstrated particularly consistent benefits across activities, whereas motion-aware contextual information showed more selective effects depending on ambulatory movement characteristics. These findings highlight the importance of contextual wearable sensing strategies for scalable calibration-free ambulatory physiological monitoring.