Heart Failure Prediction using Modal Decomposition and Masked Autoencoders for Scarce Echocardiography Databases
Andrés Bell-Navas, María Villalba-Orero, Enrique Lara-Pezzi, Jesús Garicano-Mena, Soledad Le Clainche · 2025
Read Heart Failure Prediction using Modal Decomposition and Masked Autoencoders for Scarce Echocardiography Databases by Andrés Bell-Navas, María Villalba-Orero, Enrique Lara-Pezzi, Jesús Garicano-Mena, Soledad Le Clainche (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-07606.txt.
Description
Andrés Bell-Navas and co-authors predict not just whether a heart will fail, but when, from ultrasound videos of the beating heart. Echocardiography databases are usually too small to train deep networks well, which blocks such forecasts. Their 2025 system attacks the data problem first, using a modal decomposition method to multiply and mine the video frames, turning each study into many annotated images rich in motion features.
A vision transformer then learns from those images under a masked autoencoder scheme, mixing self-supervised pretraining with supervised fine-tuning so it can train from scratch on scarce data. The network reads echo sequences in real time and estimates the time to failure. Across tests it beat established transformer and convolutional rivals. The authors fold the code into their open ModelFLOWs software for others to use.
Catalog record
| CREATOR | Andrés Bell-Navas, María Villalba-Orero, Enrique Lara-Pezzi, Jesús Garicano-Mena, Soledad Le Clainche |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 89 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-07606