Attention-based U-net: Joint Segmentation of Layers and Fluids from Retinal OCT Images

Melinščak, Martina

doi:10.23919/mipro57284.2023.10159914

Cited by 4 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike the conventional U-Net framework, which processes all regions of an input image indiscriminately, the Attention U-Net introduces attention gates. These gates implement a selective prioritization strategy, focusing on areas containing the left ventricle, thus improving segmentation accuracy [27], [30]This focused approach is particularly beneficial when image clarity is compromised or when the left ventricle's boundaries are not clearly visible. Additionally, it enhances the model's ability to accurately define complex cardiac structures.…”

Section: Resultsmentioning

confidence: 99%

Revolutionizing Echocardiography: A Comparative Study of Advanced AI Models for Precise Left Ventricular Segmentation

Kim,

Chae,

Lee

2024

Int. J. Adv. Sci. Eng. Inf. Technol.

View full text Add to dashboard Cite

Cardiovascular diseases, a leading cause of global mortality, underscore the urgency for refined diagnostic techniques. Among these, cardiomyopathies characterized by abnormal heart wall thickening present a formidable challenge, exacerbated by aging populations and the side effects of chemotherapy. Traditional echocardiogram analysis, demanding considerable time and expertise, now faces overwhelming pressure due to escalating demands for cardiac care. This study addresses these challenges by harnessing the potential of Convolutional Neural Networks, specifically YOLOv8, U-Net, and Attention U-Net, leveraging the EchoNet-Dynamic dataset from Stanford University Hospital to segment echocardiographic images. Our investigation aimed to optimize and compare these models for segmenting the left ventricle in echocardiography images, a crucial step for quantifying key cardiac parameters. We demonstrate the superiority of U-Net and Attention U-Net over YOLOv8, with Attention U-Net achieving the highest Dice Coefficient Score due to its focus on relevant features via attention mechanisms. This finding highlights the importance of model specificity in medical image segmentation and points to attention mechanisms. The integration of AI in echocardiography represents a pivotal shift toward precision medicine, improving diagnostic accuracy and operational efficiency. Our results advocate for the continued development and application of AI-driven models, underscoring their potential to transform cardiovascular diagnostics through enhanced precision and multimodal data integration. This study validates the effectiveness of state-of-the-art AI models in cardiac function assessment and paves the way for their implementation in clinical settings, thereby contributing significantly to the advancement of cardiac healthcare delivery.

show abstract

Section: Resultsmentioning

confidence: 99%

Revolutionizing Echocardiography: A Comparative Study of Advanced AI Models for Precise Left Ventricular Segmentation

Kim,

Chae,

Lee

2024

Int. J. Adv. Sci. Eng. Inf. Technol.

View full text Add to dashboard Cite

show abstract

“…Melinščak 45 employed an attention‐based U‐Net model to perform the segmentation of retinal layers and retinal fluids using the AROI dataset. The achieved dice scores for PED, SRF, and IRF were 0.674, 0.600, and 0.563, respectively.…”

Section: Resultsmentioning

confidence: 99%

Application program interface for automatic segmentation of retinal layers and fluids in Optical Coherence Tomography ‐ Neovascular Age related Macular degeneration retinal images using deep learning models

Jeya Prabha,

Sameera Fathimal,

Meghana

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Optical coherence tomography is a non‐invasive imaging technique that provides micrometer‐resolution images of retinal structures. These images can assist in identifying changes under the retina's surface, such as edema. This study proposes a novel deep learning model AR U‐Net++ for segmenting retinal layers and fluids. The four retinal layers ILM (Internal Limiting Membrane), IPL (Inner Plexiform Layer), RPE (Retinal Pigment Epithelium), BM (Bruch Membrane), and IRF (Intra Retinal Fluid), SRF (Sub Retinal Fluid), and PED (Pigment Epithelial Detachment) are segmented using AR U‐Net++. The proposed architecture AR U‐Net++ achieves better accuracy (99.67%), mean IoU (0.84), and dice coefficient (0.94) than the existing models of U‐Net, AR U‐Net, and AR W‐Net. The novelty of the suggested model AR U‐Net++ is to identify the exact location and depth of the retinal fluid in between the retinal layers and generating reports that aids the clinicians in the diagnosis of Age related Macular Degeneration.

show abstract

“…Attention mechanisms empower models to selectively focus on crucial features or regions, enhancing performance in tasks like image segmentation. Various studies have suggested incorporating attention blocks or modules into the U-Net’s encoder or decoder [ 35 , 36 , 37 , 38 ]. These attention mechanisms may utilise techniques such as channel attention, spatial attention, or self-attention.…”

Section: Related Workmentioning

confidence: 99%

Advancing Ocular Imaging: A Hybrid Attention Mechanism-Based U-Net Model for Precise Segmentation of Sub-Retinal Layers in OCT Images

Karn,

Abdulla

2024

Bioengineering

View full text Add to dashboard Cite

This paper presents a novel U-Net model incorporating a hybrid attention mechanism for automating the segmentation of sub-retinal layers in Optical Coherence Tomography (OCT) images. OCT is an ophthalmology tool that provides detailed insights into retinal structures. Manual segmentation of these layers is time-consuming and subjective, calling for automated solutions. Our proposed model combines edge and spatial attention mechanisms with the U-Net architecture to improve segmentation accuracy. By leveraging attention mechanisms, the U-Net focuses selectively on image features. Extensive evaluations using datasets demonstrate that our model outperforms existing approaches, making it a valuable tool for medical professionals. The study also highlights the model’s robustness through performance metrics such as an average Dice score of 94.99%, Adjusted Rand Index (ARI) of 97.00%, and Strength of Agreement (SOA) classifications like “Almost Perfect”, “Excellent”, and “Very Strong”. This advanced predictive model shows promise in expediting processes and enhancing the precision of ocular imaging in real-world applications.

show abstract

Attention-based U-net: Joint Segmentation of Layers and Fluids from Retinal OCT Images

Cited by 4 publications

References 18 publications

Revolutionizing Echocardiography: A Comparative Study of Advanced AI Models for Precise Left Ventricular Segmentation

Revolutionizing Echocardiography: A Comparative Study of Advanced AI Models for Precise Left Ventricular Segmentation

Application program interface for automatic segmentation of retinal layers and fluids in Optical Coherence Tomography ‐ Neovascular Age related Macular degeneration retinal images using deep learning models

Advancing Ocular Imaging: A Hybrid Attention Mechanism-Based U-Net Model for Precise Segmentation of Sub-Retinal Layers in OCT Images

Contact Info

Product

Resources

About