Ejection fraction on a budget: mapping the accuracy-compute trade space for video-based ejection fraction estimation
- 1The University of Texas at Austin
Abstract
Deep video networks estimate left ventricular ejection fraction (EF) from echocardiograms with expert-level accuracy, but the compute cost of running them is rarely reported, which leaves anyone building a handheld or bedside tool without guidance on what to deploy. We measured the accuracy versus compute trade space for EF estimation on EchoNet-Dynamic by training 22 configurations that vary clip length (8 to 64 frames), frame sampling period (1 to 4), and backbone (R(2+1)D-18, R3D-18, MC3-18, X3D-S, X3D-M, and a 2D ResNet-18 with temporal pooling), under one fixed training recipe. Every configuration was scored on accuracy (mean absolute error, R2, Bland-Altman agreement), on clinical utility (sensitivity and specificity at the EF 40% and 50% treatment thresholds, error stratified by EF band), and on cost (floating point operations, parameters, GPU and CPU latency, peak memory) under a single frozen measurement protocol. Headline claims were stress tested with replicate training seeds. Sparse temporal sampling consistently beat dense sampling: at a fixed frame count, period 4 improved mean absolute error by about one full point over period 1 across all nine cross-seed pairings while also cutting per-video cost. A plain R3D-18 achieved the best point accuracy in the study (mean absolute error 3.99), statistically tied with the reference, at 19% less CPU latency, and a 16-frame, period-4 R(2+1)D-18 halved the reference cost with no statistically confirmed accuracy loss, though with a small seed-consistent disadvantage that no single seed reveals. Removing temporal modeling entirely collapsed accuracy (mean absolute error 5.65), placing a floor under how cheap this task can get. We release the code, the cost protocol, and all per-configuration results.
Related
- Predictors of Favourable Response to Cardiac Resynchronization Therapy in Patients with Heart Failure with Reduced Ejection Fraction: A Narrative Review
- Planned Impella 5.5 Support via Brachiocephalic Artery or Ascending Aorta for Low Ejection Fraction, High-Risk Cardiac Surgery
- Physiological Basis of Skeletal Muscle Electrical Stimulation in Heart Failure with Reduced Ejection Fraction: Established Mechanisms, Extrapolated Mechanisms and Translational Priorities
- Acute-Onset Iron Overload Cardiomyopathy in Diamond-Blackfan Anaemia: A Case Report of Rapid Ejection Fraction Decline
- Presumed right atrial myxoma causing mechanical inflow obstruction and a heart-failure-with-preserved-ejection-fraction–mimicking phenotype in a resource-limited setting: A case report
BibTeX
@misc{pandey2026ejection,
title = {{Ejection fraction on a budget: mapping the accuracy-compute trade space for video-based ejection fraction estimation}},
author = {A. Pandey and K. Sharma and A. Shah},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.02.26362055},
url = {https://www.medrxiv.org/content/10.64898/2026.09.02.26362055v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Ejection fraction on a budget: mapping the accuracy-compute trade space for video-based ejection fraction estimation
AU - A. Pandey
AU - K. Sharma
AU - A. Shah
PY - 2026
DA - 2026/09/02
PB - medRxiv
DO - 10.64898/2026.09.02.26362055
UR - https://www.medrxiv.org/content/10.64898/2026.09.02.26362055v1
AB - Deep video networks estimate left ventricular ejection fraction (EF) from echocardiograms with expert-level accuracy, but the compute cost of running them is rarely reported, which leaves anyone building a handheld or bedside tool without guidance on what to deploy. We measured the accuracy versus compute trade space for EF estimation on EchoNet-Dynamic by training 22 configurations that vary clip length (8 to 64 frames), frame sampling period (1 to 4), and backbone (R(2+1)D-18, R3D-18, MC3-18, X3D-S, X3D-M, and a 2D ResNet-18 with temporal pooling), under one fixed training recipe. Every configuration was scored on accuracy (mean absolute error, R2, Bland-Altman agreement), on clinical utility (sensitivity and specificity at the EF 40% and 50% treatment thresholds, error stratified by EF band), and on cost (floating point operations, parameters, GPU and CPU latency, peak memory) under a single frozen measurement protocol. Headline claims were stress tested with replicate training seeds. Sparse temporal sampling consistently beat dense sampling: at a fixed frame count, period 4 improved mean absolute error by about one full point over period 1 across all nine cross-seed pairings while also cutting per-video cost. A plain R3D-18 achieved the best point accuracy in the study (mean absolute error 3.99), statistically tied with the reference, at 19% less CPU latency, and a 16-frame, period-4 R(2+1)D-18 halved the reference cost with no statistically confirmed accuracy loss, though with a small seed-consistent disadvantage that no single seed reveals. Removing temporal modeling entirely collapsed accuracy (mean absolute error 5.65), placing a floor under how cheap this task can get. We release the code, the cost protocol, and all per-configuration results.
ER -