Multiview Echocardiography Studies for Ventricular Abnormality Detection

30 echocardiography studies provide a minimal example dataset for multiview deep learning. Each study includes three standard non-Doppler video views and study-level labels for ventricular abnormality. Videos are preprocessed to 224x224 pixels with 64 frames, intended for code reproducibility, not clinical modeling.

Use Cases

• Load and test multiview model architectures using the three standard echocardiography views (apical four-chamber, apical two-chamber, parasternal long-axis).
• Reproduce model-loading and training code for ventricular abnormality detection from study-level labels.
• Illustrate video preprocessing steps, including cropping to the echo cone and removing burned-in text, on a small-scale dataset.

Strengths

• Contains 30 distinct echocardiography studies, providing a concrete example of multiview data structure.
• Each study includes three standard non-Doppler video views, enabling multiview algorithm testing.
• Videos are consistently preprocessed to 224x224 pixels with 64 frames, ensuring format uniformity.

Limitations

• Only 30 studies are included, which is explicitly noted as too small for clinical modeling.
• Dataset is a minimal example intended only for code reproducibility, limiting its analytical scope.
• Contains only study-level labels, lacking frame-by-frame or pixel-level annotations.

Provenance

Source

ICPSR Harvested Dataverse

Collection Method

Echocardiography studies preprocessed as described in the associated manuscript, including cropping and resizing.

Time Range

null

Freshness

null

Geography

null