Biomedical signals as time-series data generated by physiological processes: electrical, mechanical, acoustic, and chemical origins; Classification of biomedical signals: deterministic vs. stochastic, continuous vs. d...
Biomedical signals as time-series data generated by physiological processes: electrical, mechanical, acoustic, and chemical origins; Classification of biomedical signals: deterministic vs. stochastic, continuous vs. discrete, stationary vs. non-stationary; Key signal types and their clinical context: ECG (cardiac electrical activity), EEG (brain waves), EMG (muscle activity), PPG (blood volume pulse), and PCG (heart sounds); Sampling and digitization: the Nyquist criterion applied to physiological bandwidths; Signal quality metrics: SNR, baseline wander, and motion artifacts as noise modeling problems; Introduction to standard biomedical data formats: EDF, HL7, and DICOM headers for time-series records.
The Fourier Transform as the bridge between time-domain waveforms and frequency-domain physiology: interpreting spectral content of ECG and EEG; Short-Time Fourier Transform (STFT) and spectrograms for non-stationary signals; Discrete Wavelet Transform (DWT) for multi-resolution analysis: decomposing EEG into delta, theta, alpha, beta, and gamma bands; Digital filtering: FIR and IIR filter design for artifact removal (powerline interference at 50/60 Hz, baseline wander); Convolution as the computational primitive of linear filtering; Feature extraction from time-domain (mean, variance, zero-crossing rate) and frequency-domain (band power, spectral entropy) for downstream analysis.
Cardiac cycle anatomy and the PQRST complex as a structured pattern in ECG; Pan-Tompkins algorithm for QRS detection: bandpass filtering, differentiation, squaring, and moving-window integration as a classical signal processing pipeline; R-R interval time series and Heart Rate Variability (HRV) as a window into the autonomic nervous system; Seizure detection in EEG: energy-based and threshold-based detectors; Apnea detection from respiratory signals using autocorrelation and spectral analysis; The general paradigm: segmentation feature extraction event classification as a reusable computational template across all biosignal types.
Transducer principles: converting physiological quantities into electrical signals (electrodes, pressure transducers, optical sensors); Instrumentation amplifiers, differential amplification, and common-mode rejection as analog front-end design concepts; Analog-to-Digital conversion pipeline: anti-aliasing filters, ADC resolution, and dynamic range; Wearable sensor systems: IMUs, pulse oximeters, and continuous glucose monitors as constrained embedded data acquisition devices; Data streaming and buffering: circular buffers and real-time processing constraints; Introduction to IoT-based patient monitoring architectures: edge preprocessing, MQTT-based transmission, and cloud aggregation.
The clinical decision support system (CDSS) as a software pipeline: signal ingestion preprocessing feature engineering classification alert generation; Arrhythmia classification from ECG as a canonical supervised learning problem: feature-based classifiers vs. end-to-end approaches; Stress and fatigue detection from HRV and EEG features; Alarm fatigue in ICUs: the false-positive problem and precision-recall tradeoffs in safety-critical systems; Introduction to interoperability standards: FHIR as a RESTful API standard for exchanging clinical data; Regulatory and ethical constraints: FDA classification of Software as a Medical Device (SaMD) and data privacy under HIPAA.