Echocardiogram Vector Embeddings Via R3D Transformer for the Improvement of Ejection Fraction Estimation Instruments
Bibliographic record
Abstract
The estimation of ejection fraction (EF) is critical to intensive care. Doctors commonly use estimated EF to create plans for ICU patients, and a low EF can indicate ventricular systolic dysfunction, which increases the risk of adverse events including heart failure. Recently, interest has grown in deep learning (DL) instruments that can measure cardiac activity to estimate EF automatically. In particular, the vector embeddings learned by DL-based EF estimation models provide valuable mappings of echocardiograms within latent space, which can be utilized to understand the error patterns of DL-based EF estimators and thus improve these instruments. In this work, we provide those embeddings. To this end we repurpose an R3D transformer, a state-of-the-art deep learning model for Video Action Recognition, to classify whether patients have ventricular dysfunction or not (ejection fraction below or above 50%) using echocardiogram data. Our R3D model achieves a test AUC of 0.916 and a test accuracy of 87.5%, approaching the performance of previous comparable studies with a fraction of the training time. Most importantly, the vector embeddings learned by this model will enable future analysis of the errors of DL-based EF estimators, democratizing the improvement of these instruments. The quality of these embeddings, furthermore, is evidenced by the strong results of the model that learned them.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".