How akhil-preetham/cardio-diag-updated works
CARDIODIAG is a retrospective coronary artery disease research implementation, with a guarded five-feature tool for producing research scores as its clearest user-facing workflow. Its broader runtime supports dataset preparation, frozen-partition experiments, model comparison, validation and transport analysis, and evidence/report generation. The graph preserves those research stages and the sampled portable scoring path. Several experiment modules are present in the tree but were not excerpted; their connections are omitted unless documented or indexed.
Main entry points: portable_predict.py Portable scoring · primary_research_v2/run.py Primary benchmark · audit.py Phase audit · phase4_validation_audit/run.py Validation experiments
Components 21
Research experiments
- Dataset acquisition
CARDIODIAG_FRESH/runner/data/collect_clinical_datasets_v2.py - Frozen partitions
CARDIODIAG_FRESH/runner/methodology_corrections_v2/split_manifest.py - Primary benchmark
CARDIODIAG_FRESH/runner/primary_research_v2/run.py - Validation experiments
CARDIODIAG_FRESH/runner/phase4_validation_audit/run.py - ZAS selection
CARDIODIAG_FRESH/runner/phase6_zas_selection/run.py - Cohort transport
CARDIODIAG_FRESH/runner/research_extension_v1/transport.py
Evaluation and audits
- Phase audit
CARDIODIAG_FRESH/runner/phase6_1_audit/audit.py - Paired statistics
CARDIODIAG_FRESH/src/cardiodiag/statistics/paired.py - Data quality
CARDIODIAG_FRESH/src/cardiodiag/data/quality.py - Calibration metrics
CARDIODIAG_FRESH/src/cardiodiag/calibration/metrics.py - Subgroup fairness
CARDIODIAG_FRESH/src/cardiodiag/fairness/subgroups.py - LIME explanations
CARDIODIAG_FRESH/src/cardiodiag/explainability/lime_workflow.py
Portable scoring
- Portable scoring
tools/portable_predict.py - Input validation
CARDIODIAG_FRESH/runner/research_extension_v1/predict_research.py - Runtime and model gate
tools/portable_release.py
Evidence and reporting
- Evidence manuscript
research_publication_plan/build_manuscript.py - Research report
CARDIODIAG_FRESH/paper/research_extension_v1/build_report.py
Other components
- Research user
- Clinical datasets
- Scoring input
- Research scores
Connections 15
- Clinical datasets → Dataset acquisition: supplies data
- Dataset acquisition → Primary benchmark: prepares inputs
- Frozen partitions → Primary benchmark: defines partitions
CARDIODIAG_FRESH/runner/primary_research_v2/run.py - ZAS selection → Phase audit: reconciles results
CARDIODIAG_FRESH/runner/phase6_1_audit/audit.py - Validation experiments → Phase audit: supports audit
- Research user → Portable scoring: requests scoring
README.md - Scoring input → Portable scoring: provides rows
README.md - Portable scoring → Runtime and model gate: checks runtime
tools/portable_predict.py - Portable scoring → Input validation: validates and scores
tools/portable_predict.py - Input validation → Cohort transport: uses normalization
CARDIODIAG_FRESH/runner/research_extension_v1/predict_research.py - Portable scoring → Research scores: writes scores
README.md - Primary benchmark → Evidence manuscript: informs reporting
- Phase audit → Research report: informs reporting
- Evidence manuscript → Paired statistics: imports statistics interface
research_publication_plan/build_manuscript.py - Validation experiments → Paired statistics: uses comparisons