Related Experiment Video
Updated: May 31, 2025
12:22
Multimodal Volumetric Retinal Imaging by Oblique Scanning Laser Ophthalmoscopy oSLO and Optical Coherence Tomography OCT
Published on: August 4, 2018
8.5K
An Efficient Retinal Fluid Segmentation Network Based on Large Receptive Field Context Capture for Optical Coherence
Hang Qi1, Weijiang Wang1,2, Hua Dang1
1School of Integrated Circuits and Electronics, Beijing Institute of Technology, Beijing 100081, China.
Entropy (Basel, Switzerland)
|January 24, 2025
Summary
This study introduces LKMU-Lite, a novel lightweight deep learning model for segmenting retinal fluids in Optical Coherence Tomography (OCT) images. The efficient method achieves state-of-the-art accuracy, overcoming common challenges in retinal imaging analysis.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Optical Coherence Tomography (OCT) is vital for diagnosing retinal diseases.
- Accurate segmentation of retinal fluid and lesions in OCT images is challenging due to image artifacts.
- Existing methods often involve high computational costs for effective feature extraction.
Purpose of the Study:
- To develop an efficient and accurate deep learning model for retinal fluid segmentation in OCT images.
- To address the limitations of computational overhead in current segmentation techniques.
- To improve the detection of various fluid and lesion types in retinal scans.
Main Methods:
- Proposed LKMU-Lite, a lightweight U-shaped network for retinal fluid segmentation.
- Integrated Decoupled Large Kernel Attention (DLKA) for enhanced local and long-range feature representation.
- Employed Multi-scale Group Perception (MSGP) with dilated convolutions for diverse lesion size detection.
- Introduced an Aggregating-Shift decoder to reduce complexity while maintaining feature integrity.
Main Results:
- LKMU-Lite achieved state-of-the-art performance on ICF and RETOUCH datasets.
- The model demonstrated superior accuracy across multiple segmentation metrics.
- Achieved high efficiency with only 1.02 million parameters and 3.82 G FLOPs.
Conclusions:
- LKMU-Lite offers an efficient and effective solution for retinal fluid segmentation in OCT imaging.
- The proposed architecture successfully balances model complexity and segmentation performance.
- Demonstrated generalizability and state-of-the-art results compared to existing methods.
