Deep Neural Network Regression for Automated Retinal Layer Segmentation in Optical Coherence Tomography Images

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Accurate segmentation of retinal layers in OCT images is crucial for diagnosing eye diseases and preventing blindness.
• Current manual segmentation methods are time-consuming and prone to human bias.
• There is a need for automated, robust, and efficient OCT image segmentation techniques.

Purpose of the Study:

• To develop an automated, bias-free method for segmenting retinal layers in OCT images.
• To improve the accuracy and efficiency of retinal layer quantification for early disease detection.
• To address the limitations of manual segmentation in clinical practice.

Main Methods:

• Proposed a deep neural network regression model for automated OCT image segmentation.
• Utilized image intensity, gradient, and adaptive normalized intensity score (ANIS) as features for learning.
• Reformulated segmentation as a regression problem to reduce dataset requirements and complexity.

Main Results:

• Achieved high accuracy with a Dice similarity coefficient of approximately 0.966.
• Demonstrated robustness to image variations like low contrast, noise, and blood vessels, assisted by ANIS.
• Processed images efficiently, with an average time of 10.596 seconds per image for eight boundary identifications.

Conclusions:

• The proposed automated segmentation method is accurate, time-efficient, and robust for OCT images.
• This technique can significantly aid in the early diagnosis of retinal diseases and prevention of vision loss.
• The regression-based approach offers a practical solution for automated retinal layer segmentation.