Holter ECG Analytical System
A complete analytical engine for the first level of expert review of long-term ambulatory ECG recordings.
The system was designed to transform hours or days of noisy physiological data into structured clinical information before physician review — while keeping the final diagnosis and all clinical decisions with the physician.
From raw recording to usable evidence.
The problem was not to build another ECG pre-processing tool. It was to perform the first level of expert analysis before the physician began diagnostic review.
Manual inspection of long Holter recordings is dominated by routine work: finding usable segments, locating beats, measuring intervals, classifying events and navigating to episodes that may matter clinically. The analytical core was built to perform that work systematically and present the result as structured evidence.
My contribution focused on the analytical engine itself: signal-processing architecture, detection and classification algorithms, decision logic, and the research platform used to develop, validate and refine the pipeline.
A modular analytical pipeline.
Each stage solves a specific task and prepares reliable information for the next stage. This makes the system easier to validate, debug, extend and evolve module by module.
Different physiological events need different models.
The system was not designed as one monolithic detector. Heartbeat detection, ECG delineation, interval measurement, beat classification and rhythm analysis were treated as distinct physical problems.
Heartbeat detection was formulated as identification of the characteristic event corresponding to ventricular contraction rather than simple peak picking. Subsequent modules then analyzed the individual ECG components — P wave, QRS complex, T wave and U wave — using criteria appropriate to each component.
Structured clinical information, not just signal marks.
The pipeline was designed to produce the material a physician needs for review: signal-quality assessment, heartbeat and rhythm classification, interval measurements, arrhythmia detection, statistical summaries, representative ECG episodes and a preliminary diagnostic report with direct navigation to significant events.
Engineering platform.
The system was implemented as a modular software platform intended for efficient processing of long-term Holter recordings and independent development and validation of individual analytical modules.