ReLayNet: retinal layer and fluid segmentation of macular optical coherence tomography using fully convolutional...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic macular edema diagnosis relies on non-invasive Optical Coherence Tomography (OCT) to assess retinal layers.
• Accurate segmentation of retinal layers and fluid is crucial for effective diagnosis and treatment monitoring.

Purpose of the Study:

• To propose ReLayNet, a novel fully convolutional deep architecture for automated, end-to-end segmentation of retinal layers and fluid masses in OCT scans.
• To evaluate ReLayNet's effectiveness against state-of-the-art segmentation methods.

Main Methods:

• Developed ReLayNet, a deep architecture with encoder-decoder paths for semantic segmentation of OCT images.
• Trained ReLayNet using a joint loss function combining weighted logistic regression and Dice overlap loss.
• Validated the framework on a public benchmark dataset.

Main Results:

• ReLayNet demonstrated effectiveness in segmenting retinal layers and fluid masses.
• Performance was compared against five state-of-the-art segmentation methods, including deep learning approaches.
• The proposed method showed competitive or superior results in substantiating its effectiveness.

Conclusions:

• ReLayNet provides an effective deep learning solution for automated retinal layer and fluid segmentation in OCT scans.
• This advancement can aid in the non-invasive diagnosis and assessment of diabetic macular edema.
• The study highlights the potential of deep learning in improving ophthalmic image analysis.