HFE-Mamba: High-Frequency Enhanced Mamba Network for Efficient Segmentation of Left Ventricle in Pediatric Echocardiograms
Open Access
- 8 July 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 123038-123048
Abstract
Automated ventricular function analysis can enhance healthcare consistency and accessibility, particularly in resource-limited settings. Current segmentation methods trained on adult heart ultrasounds struggle to accurately outline the irregular shape of the left ventricle owing to their limited exploration of border features. HFE-Mamba is introduced for left ventricle segmentation with shape awareness in order to address the existing challenge. Therefore, this proposal introduces the High-Frequency Enhancement Block (HFEB), enhancing the high-frequency component of left ventricles, particularly the boundary area in pediatric echocardiograms. Moreover, this also facilitates the investigation of target boundary specifics while extracting features. The incorporation of newly suggested vision mamba layers into encoder and decoder branches enhances the model’s computational and memory efficiency while capturing global dependencies. Tests conducted on two publicly available datasets indicate the superior predictive accuracy of the HFE-Mamba model in identifying target shapes.Keywords
This publication has 27 references indexed in Scilit:
- EchoEFNet: Multi-task deep learning network for automatic calculation of left ventricular ejection fraction in 2D echocardiographyComputers in Biology and Medicine, 2023
- Video-Based Deep Learning for Automated Assessment of Left Ventricular Ejection Fraction in Pediatric PatientsJournal of the American Society of Echocardiography, 2023
- MaxViT: Multi-axis Vision Transformer2022
- MAEF-Net: Multi-attention efficient feature fusion network for left ventricular segmentation and quantitative analysis in two-dimensional echocardiographyUltrasonics, 2022
- Contrastive Pretraining for Echocardiography Segmentation with Limited Data2022
- Multi-scale wavelet network algorithm for pediatric echocardiographic segmentation via hierarchical feature guided fusionApplied Soft Computing, 2021
- Dual attention enhancement feature fusion network for segmentation and quantitative analysis of paediatric echocardiographyMedical Image Analysis, 2021
- Convolutional-Neural-Network-Based Approach for Segmentation of Apical Four-Chamber View from Fetal EchocardiographyIEEE Access, 2020
- AIDAN: An Attention-Guided Dual-Path Network for Pediatric Echocardiography SegmentationIEEE Access, 2020
- Birth Prevalence of Congenital Heart Disease WorldwideJACC, 2011