فهرست نشریات علمی دسترسی آزاد ایران

The convergence of the Internet of Medical Things (IoMT), biosignal processing and artificial intelligence has redefined early disease detection by enabling real-time, remote and ultra-low-power analytics at the edge. In this systematic review we analyse 40 peer-reviewed studies (2020–2025) that apply machine-learning or deep-learning techniques to electrocardiogram (ECG), electroencephalogram (EEG), electromyogram (EMG) and electrooculogram (EOG) data within IoMT ecosystems. We chronicle the complete pipeline—from sensor physics (AD8232, Myoware, Muse 2) through Fog/Edge preprocessing, feature engineering and model selection, to deployment on Raspberry Pi, ESP32 or Jetson Nano boards. Reported clinical accuracies peak at 99 % for arrhythmia, 98.5 % for neonatal seizure and 97 % for stress detection, while end-to-end latency is pushed below 120 ms and raw-data uploads are reduced by 99 % via on-device inference. Methodological trends show a migration from classical SVM/Random-Forest-PSO models (2020) to compressed CNN-LSTM-GRU hybrids (2025) and multimodal CNN-Transformer architectures. Remaining challenges include patient-specific fine-tuning, scalable 5G-Edge orchestration, multimodal fusion and zero-trust security. The evidence indicates that hybrid Edge-Fog-Cloud frameworks which couple pruned deep nets with optimised classical learners currently offer the most pragmatic route toward large-scale, privacy-preserving and clinically validated IoMT diagnostics.