Automated Layer Segmentation of Retinal Optical Coherence Tomography Images Using a Deep Feature Enhanced Structured...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Optical coherence tomography (OCT) is vital for ophthalmic diagnosis.
• Manual retinal layer segmentation is time-consuming and subjective.
• Accurate segmentation is crucial for diagnosing retinal diseases.

Purpose of the Study:

• To develop an automated method for retinal OCT image segmentation.
• To improve the accuracy and efficiency of retinal layer segmentation.
• To provide a reliable tool for ophthalmic diagnosis and research.

Main Methods:

• Integration of deep convolutional features (from deep residual network) and hand-designed features.
• Training a structured random forests classifier.
• Utilizing contour probability graphs and shortest path algorithm for final segmentation.

Main Results:

• Achieved a mean layer contour error of 1.215 pixels, outperforming the state-of-the-art (1.464 pixels).
• Obtained an F1-score of 0.885, surpassing the state-of-the-art (0.863).
• Demonstrated superior performance in automated retinal layer segmentation.

Conclusions:

• The proposed method offers an effective and accurate solution for automated retinal OCT image segmentation.
• This technique can significantly aid clinicians in diagnosing and studying retinal diseases.
• The integration of deep and hand-designed features shows promise for medical image analysis.