Sarcopenia Assessment Using a Bioimpedance-Based Approach

Sarcopenia assessment using a bioimpedance-based approach integrates electrical impedance myography (EIM) with advanced machine learning algorithms to estimate skeletal muscle fat-free mass from localized muscle impedance measurements. Consistent with Prof. Cheng’s research in biomedical instrumentation and impedance signal modeling, this approach emphasizes feature extraction from multi-frequency impedance parameters and regression-based predictive modeling to enhance estimation accuracy. By replacing conventional, time-consuming physical performance tests with quantitative impedance-derived biomarkers, the proposed system improves objectivity and clinical feasibility. Furthermore, the development of a portable and low-cost bioimpedance measurement device aligns with translational biomedical engineering principles, enabling point-of-care sarcopenia screening and longitudinal muscle health monitoring.

Electrical Impedance Tomography (EIT) System Development and Applications

Electrical Impedance Tomography (EIT) is a noninvasive imaging modality capable of monitoring cardiopulmonary function through dynamic impedance distribution analysis. Building upon Prof. Cheng’s work in EIT system development and image reconstruction algorithms, this study proposes a deep learning–based framework for separating cardiac and pulmonary impedance components from time-series EIT data. The integration of convolutional neural networks with temporal modeling enhances spatial resolution and dynamic interpretability. In addition, machine learning–based texture and morphological feature extraction methods are applied to muscle impedance images for quantitative characterization. These improvements in reconstruction stability and image quality facilitate more reliable muscle condition evaluation and may further support impedance-based sarcopenia assessment.

Obstructive Sleep Apnea Phenotyping

For obstructive sleep apnea (OSA) phenotyping, the proposed framework combines nasal pressure waveform analysis with nonlinear dynamic feature extraction using the Weighted Visibility Graph (WVG) method. In accordance with Prof. Cheng’s expertise in biomedical signal processing and intelligent classification systems, support vector machine (SVM)-based feature selection is employed to identify discriminative physiological markers. A Fuzzy Inference System (FIS) is subsequently constructed to predict postoperative apnea–hypopnea index (AHI), offering both predictive robustness and physiological interpretability. This hybrid machine learning architecture enhances model transparency while maintaining high classification performance, supporting precision medicine strategies in sleep disorder management.

Histopathologic Section Analysis Using Whole Slide Imaging

Whole slide imaging (WSI) has become a fundamental platform in digital pathology, enabling high-resolution visualization of histopathologic sections. However, due to gigapixel-scale image sizes across multiple magnifications, computational efficiency remains a significant challenge. Reflecting Prof. Cheng’s contributions in medical image processing and system optimization, the proposed framework employs hierarchical patch-based analysis combined with deep convolutional neural networks to reduce computational burden while preserving diagnostic information. Multi-scale feature fusion and automated region-of-interest detection further improve analytical efficiency. Such system-level optimization supports scalable and intelligent histopathological assessment, facilitating precision diagnostics in clinical workflows.