How rajb-1003/ecg-classification works
ECG Classifier 2.0 turns uploaded or dataset ECG recordings into quality-screened, classified and explainable clinical reports. The principal inference workflow is initiated by the CLI or Streamlit UI and proceeds through loading, cleaning, quality assessment, waveform/feature analysis, classification and reporting. The repository also contains a separate configurable dataset-processing and signal-quality pipeline, plus model training and validation capabilities. Edges follow the sampled entry-point references and README workflow; unsampled internals are represented only where documented.
Main entry points: main.py CLI entry point · app.py Streamlit UI · loaders.py Dataset loaders · run_pipeline.py Quality pipeline
Components 29
ECG inference
- CLI entry point
main.py - Streamlit UI
app.py - ECG input
- Record loading
ecg_loader.py
Signal analysis
- ECG cleaning
preprocessing.py - Signal quality
signal_quality.py - PQRST delineation
delineation.py - Feature extraction
feature_extraction.py
Clinical interpretation
- Conduction analysis
conduction_analysis.py - Rhythm classification
rhythm_classifier.py - Decision fusion
decision_fusion.py - Clinical rules
clinical_rules.py - SHAP explainability
explainability.py - Clinical reports
report_generator.py - ECG visualizations
ecg_visualizer.py
Quality pipeline
- Dataset loaders
core/dataset_loaders/loaders.py - Production preprocessing
core/preprocessing/preprocessor.py - Hardware sanity gate
core/hardware_gate/gate.py - SQA classifier
core/sqa/classifier.py - Quality features
core/feature_extraction/extractor.py - Quality reporting
core/reporting/generator.py - Filter configuration
config/filter_config.py - Dataset paths
config/paths.py - Quality pipeline
pipelines/run_pipeline.py
Models and validation
- Model training
train.py - Trained models
- Model evaluation
evaluate.py - Dataset validation
pipelines/run_batch_validation.py - External validation
pipelines/run_external_validation.py
Connections 48
- ECG input → CLI entry point: provides records
- ECG input → Streamlit UI: uploads ECGs
- CLI entry point → Record loading: loads records
main.py - CLI entry point → ECG cleaning: cleans signals
main.py - CLI entry point → Signal quality: assesses quality
main.py - CLI entry point → PQRST delineation: detects waves
main.py - CLI entry point → Feature extraction: extracts features
main.py - CLI entry point → Conduction analysis: analyzes conduction
main.py - CLI entry point → Decision fusion: fuses decisions
main.py - CLI entry point → SHAP explainability: explains predictions
main.py - CLI entry point → Clinical reports: generates reports
main.py - CLI entry point → ECG visualizations: renders plots
- Streamlit UI → Record loading: loads records
app.py - Streamlit UI → ECG cleaning: cleans signals
app.py - Streamlit UI → Signal quality: assesses quality
app.py - Streamlit UI → PQRST delineation: shows delineation
app.py - Streamlit UI → Conduction analysis: analyzes conduction
app.py - Streamlit UI → Decision fusion: classifies ECG
app.py - Streamlit UI → SHAP explainability: shows explanations
app.py - Streamlit UI → Clinical reports: shows report
- Model training → Trained models: saves models
- CLI entry point → Model training: starts training
main.py - CLI entry point → Trained models: loads models
- Model evaluation → Trained models: evaluates models
- Decision fusion → Clinical rules: applies rules
- Quality pipeline → Dataset loaders: loads datasets
pipelines/run_pipeline.py - Quality pipeline → Production preprocessing: filters signals
pipelines/run_pipeline.py - Quality pipeline → Hardware sanity gate: checks hardware
pipelines/run_pipeline.py - Quality pipeline → Quality features: extracts features
pipelines/run_pipeline.py - Quality pipeline → Quality reporting: generates reports
pipelines/run_pipeline.py - Dataset loaders → Dataset paths: uses paths
core/dataset_loaders/loaders.py - Dataset loaders → Filter configuration: reads settings
core/dataset_loaders/loaders.py - Production preprocessing → Filter configuration: reads settings
core/preprocessing/preprocessor.py - Production preprocessing → Dataset loaders: accepts records
core/preprocessing/preprocessor.py - Quality reporting → Dataset loaders: uses records
core/reporting/generator.py - Quality reporting → Hardware sanity gate: uses gate results
core/reporting/generator.py - Quality reporting → Quality features: uses features
core/reporting/generator.py - Quality reporting → Filter configuration: reads settings
core/reporting/generator.py - Dataset validation → Dataset loaders: loads datasets
pipelines/run_batch_validation.py - Dataset validation → Production preprocessing: filters signals
pipelines/run_batch_validation.py - Dataset validation → Hardware sanity gate: checks hardware
pipelines/run_batch_validation.py - Dataset validation → Quality features: extracts features
pipelines/run_batch_validation.py - Dataset validation → SQA classifier: classifies quality
pipelines/run_batch_validation.py - External validation → Dataset loaders: loads datasets
pipelines/run_external_validation.py - External validation → Production preprocessing: filters signals
pipelines/run_external_validation.py - External validation → Hardware sanity gate: checks hardware
pipelines/run_external_validation.py - External validation → Quality features: extracts features
pipelines/run_external_validation.py - External validation → SQA classifier: classifies quality
pipelines/run_external_validation.py