1 Definition 2 Functional Requirements 3 Syntax 4 Semantics

1 Definition

The ECG Processing Type defines the allowable operations, methods, and processing descriptors used for Electrocardiography (ECG) data.
It reuses Common Definitions for: Header, Algorithm, Algorithms, FeatureClass, Features

2 Functional Requirements

The ECG Processing Type shall:

  • Fix Domain = ECG.
  • Validate Operation against ECG‑specific enumerations.
  • Validate Method against ECG signal‑processing and ML techniques.
  • Allow Algorithm to be either a string identifier or an AlgorithmObject.
  • Allow Algorithms to be an array of Algorithm items.
  • Support FeatureClass to specify the type of ECG features extracted.
  • Require Features to be a non‑empty set of unique feature names.
  • Support Targets to specify the ECG waveform components (P, QRS, T, ST) involved.

3 Syntax

https://schemas.mpai.community/AIH1/V1.0/data/ECGProcessingType.json

4 Semantics

Label Description
Header ECG Processing Type Header, Standard “AIH‑CPT‑Vx.y”.
Domain Constant value "ECG". Applies exclusively to electrocardiography data. No other value allowed.
Operation Specifies the ECG‑specific processing step. Enumerated list includes: RPeakDetection, QRSDetection, QTMeasurement, STDeviation, HRVTimeDomain, HRVFrequencyDomain, BeatClassification, BaselineWanderRemoval, Denoising, ArrhythmiaDetection.
RPeakDetection Detects R‑peaks in the ECG waveform.
QRSDetection Identifies the QRS complex boundaries.
QTMeasurement Measures QT interval duration.
STDeviation Detects elevation/depression in the ST segment.
HRVTimeDomain Computes time‑domain HRV metrics (e.g., SDNN, RMSSD).
HRVFrequencyDomain Computes frequency‑domain HRV metrics (e.g., LF/HF).
BeatClassification Categorizes beats (e.g., normal, PVC, PAC).
BaselineWanderRemoval Removes low‑frequency drift (baseline wander).
Denoising Removes noise components while preserving ECG morphology.
ArrhythmiaDetection Identifies arrhythmias based on beat patterns and intervals.
Method Processing technique used to implement the operation. Must be one of: PanTompkins, WaveletTransform, SavitzkyGolay, AdaptiveThresholding, TemplateMatching, LombScargle, WelchPSD, MedianFilter, CNN, RNN, Transformer.
PanTompkins Classic QRS/R‑peak detection algorithm.
WaveletTransform Multi‑resolution analysis for feature extraction or denoising.
SavitzkyGolay Smoothing filter preserving ECG morphology.
AdaptiveThresholding Detection based on adaptive amplitude thresholds.
TemplateMatching Beat or pattern detection using template similarity.
LombScargle Periodogram for unevenly sampled HRV frequency analysis.
WelchPSD Welch method for spectral density estimation.
MedianFilter Non‑linear filtering to remove baseline drift or spikes.
CNN Convolutional neural network for classification or feature learning.
RNN Recurrent model (e.g., LSTM/GRU) for sequential ECG analysis.
Transformer Attention‑based deep‑learning architecture for ECG tasks.
Algorithm Either a string identifier or an AlgorithmObject from CommonDefinitions.
AlgorithmObject.Name Required name of the algorithm (e.g., “PanTompkins‑v2”).
AlgorithmObject.Version Optional version string.
AlgorithmObject.Params Free‑form parameter object for algorithm configuration.
Algorithms Array of Algorithm entries (string ID or AlgorithmObject).
FeatureClass Category of ECG features (morphological, temporal, HRV, spectral, etc.).
Features Non‑empty array of unique ECG feature names (e.g., RR_interval, QRS_width, ST_slope).
Targets ECG waveform components associated with the operation. One or more of: P, QRS, T, ST.
Trace Provenance information and Tine.
DescrMetadata Descriptive Metadata