Echo-POSED: Geometric Self-Distillation for Echocardiography Guidance
It addresses the need for automated echocardiography guidance to reduce operator dependency, but the results are limited to simulations and specific datasets.
Echo-POSED is a self-supervised framework for real-time TTE guidance that recommends probe adjustments from 2D ultrasound images without expert labels, achieving a combined mean angular error of 8.2 degrees in intra-patient simulations.
We introduce Echo-POSED, a self-supervised framework for real-time transthoracic echocardiography (TTE) guidance that recommends probe adjustments directly from 2D ultrasound images, without the need for expert-labelled views or tracked probe trajectories. Instead, it trains on 2D views sliced from routinely acquired 3D echocardiography volumes, enforcing equivariance to probe motions while remaining invariant to cardiac phase, yielding a pose representation on $\mathrm{SO}(3)\times\mathrm{SO}(3)$. Across a held-out split and public external 3D--TTE datasets (including vendor shift), Echo-POSED maintains geometric consistency under virtual perturbations and enables intra- and inter-patient guidance simulations, achieving a combined mean angular error of 8.2 degrees between the guided and target views in intra-patient simulations with cardiac motion.