CMOS: Confidence-Guided Multi-Scale Semi-Supervised Segmentation for Retinal Layers in OCT Images.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate segmentation of retinal layers in optical coherence tomography (OCT) images is vital for diagnosing ophthalmic diseases.
• Current methods struggle with limited annotated data and image quality variations, especially in diseased retinas.

Purpose of the Study:

• To introduce a novel semi-supervised learning method for enhanced OCT retinal layer segmentation.
• To improve segmentation accuracy and model robustness in the presence of lesions and imaging inconsistencies.

Main Methods:

• Proposed the Confidence-Guided Multi-Scale OCT Segmentation (CMOS) method.
• Incorporated bi-directional feature alignment to refine pseudo-labels using unlabeled data.
• Utilized a multi-scale aggregation (MSA) module to handle feature variability and image quality fluctuations.

Main Results:

• The CMOS method significantly enhances OCT retinal layer segmentation accuracy.
• Demonstrated superior performance compared to existing semi-supervised learning approaches.
• Showcased improved model robustness under complex pathological conditions and varying image quality.

Conclusions:

• The proposed CMOS method effectively addresses key challenges in OCT retinal layer segmentation.
• Leveraging unlabeled data and multi-scale features leads to state-of-the-art performance.
• This approach holds significant potential for quantitative analysis and diagnosis in ophthalmology.