aiqbench analysis
This study empirically evaluates how data augmentation strategies impact the cross-dataset generalizability of deep learning models for echocardiography segmentation. The research demonstrates that geometry-based transformations and complementary pairwise combinations significantly improve transfer performance, whereas aggressive intensity transforms often degrade results in domain-shift scenarios. These findings offer concrete guidance for engineers seeking to enhance model robustness when deploying AI tools across diverse medical imaging environments without access to massive new datasets.
EAGT: Echocardiography Augmentation for Generalisability and Transferability
arXiv:2605.16427v3 Announce Type: replace-cross Abstract: Deep learning models for echocardiography segmentation often struggle to generalise across institutions, scanners, and patient populations, where collecting large, consistently annotated datasets is infeasible. Data augmentation is inexpensive and widely used to improve the robustness of deep learning models; however, its role in enhancing cross-dataset generalisability in echocardiography remains insufficiently understood. This study presents a large-scale multi-dataset evaluation of 29 data augmentation techniques and their pairwise combinations for 2D left ventricular segmentation using a U-Net trained on Unity, CAMUS, and EchoNet Dynamic datasets. Each augmentation was explored under several hyperparameter settings and assessed through repeated runs using Dice and IoU in both in-domain and cross-dataset scenarios, with statistical significance quantified via independent t-tests. In-domain accuracy was near-saturated and insensitive to augmentation, whereas cross-dataset performance varied widely. Geometry-based augmentations including affine, shift-scale-rotate, flip, and perspective produced the largest and most consistent gains, while aggressive intensity- and artefact-based transforms often degraded transfer. Moreover, pairwise combinations outperformed individual augmentations mainly when the two transformations were complementary, particularly by improving some difficult domain-shift cases from poor to acceptable performance. These findings provide empirical guidance for designing augmentation policies that improve the robustness and transferability of echocardiography segmentation models.
Abstract:Deep learning models for echocardiography segmentation often struggle to generalise across institutions, scanners, and patient populations, where collecting large, consistently annotated datasets is infeasible. Data augmentation is inexpensive and widely used to improve the robustness of deep learning models; however, its role in enhancing cross-dataset generalisability in echocardiography remains insufficiently understood. This study presents a large-scale multi-dataset evaluation of 29 data augmentation techniques and their pairwise combinations for 2D left ventricular segmentation using a U-Net trained on Unity, CAMUS, and EchoNet Dynamic datasets. Each augmentation was explored under several hyperparameter settings and assessed through repeated runs using Dice and IoU in both in-domain and cross-dataset scenarios, with statistical significance quantified via independent t-tests. In-domain accuracy was near-saturated and insensitive to augmentation, whereas cross-dataset performance varied widely. Geometry-based augmentations including affine, shift-scale-rotate, flip, and perspective produced the largest and most consistent gains, while aggressive intensity- and artefact-based transforms often degraded transfer. Moreover, pairwise combinations outperformed individual augmentations mainly when the two transformations were complementary, particularly by improving some difficult domain-shift cases from poor to acceptable performance. These findings provide empirical guidance for designing augmentation policies that improve the robustness and transferability of echocardiography segmentation models.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2605.16427 [cs.CV]
(or arXiv:2605.16427v3 [cs.CV] for this version)
Full text under the publisher's open licence · 211 words
Related coverage
The Impact of Semantic Pairs on Self-Supervised Representation Learning
arXiv · Sep 22, 2026
From Deterministic to Generative Deep Learning for Urban Air Quality Reconstruction from Sparse Observations
arXiv · Jul 29, 2026
Michael Biercuk, The Q-CTRL Founder Whose Error Suppression Software Runs on IBM, Rigetti and Diraq Machines
Quantum Zeitgeist · Aug 30, 2026
Data bottlenecks won’t prevent an intelligence explosion (but they will slow it down)
LessWrong · Sep 9, 2026