What happened

DROID-RV was trained on over 380,000 clips and tested internally, in primary care and on an external Hungarian cohort.

Why it matters

Right-ventricle assessment is time-consuming and expert-dependent, creating a plausible assistive use case.

Key numbers & facts

  • 380,000+ clips
  • About 17,000 training patients
  • 9,100 primary-care tests
  • 831-patient external dataset

Practical implications

  • Evaluate calibration, subgroup performance and prospective reader impact before clinical deployment.

Likely future impact

Analysis, not a confirmed outcome.

  • Validated automation could triage abnormal studies and expand access where MRI or expert readers are scarce.

Evidence

  • The institutional summary links the peer-reviewed Nature Cardiovascular Research paper and lists conflicts.

What remains uncertain

  • Retrospective association with outcomes is not proof that using the model improves care; no approval is claimed.