CFD-Derived Radiomics for Left Atrial Flow Assessment in Atrial Fibrillation

Thirty subjects (10 controls, 20 AF patients) underwent cardiac CT for this proof-of-concept study. Lida Alinezhad published this dataset on figshare in June 2026, evaluating radiomics-based quantification of spatial texture patterns from hemodynamic maps to provide reproducible flow biomarkers for AF classification.

Use Cases

• Train classifiers for atrial fibrillation detection based on radiomics features from hemodynamic maps.
• Identify correlations between flow biomarkers and clinical markers like LA enlargement or mitral regurgitation.
• Validate the stability of radiomics features for objective hemodynamic phenotyping.
• Develop interpretable machine learning models for cardiac flow patterns using SHAP analysis.

Strengths

• Dataset includes 30 subjects with patient-specific CFD simulations.
• Five robust radiomics features were identified with high selection frequency and bootstrap stability.
• The derived classifier achieved 93% accuracy and an AUC of 0.92.
• Features showed significant correlations (ρ = 0.45–0.69) with established markers of atrial remodeling.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Hemodynamic parameter maps were generated from patient-specific CFD simulations based on contrast-enhanced cardiac CT scans.

Freshness

Last updated 2026-06-04 04:27:35; freshness should be verified.