Deep Learning Based Sub-Retinal Fluid Segmentation in Central Serous Chorioretinopathy Optical Coherence Tomography...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Automated segmentation of sub-retinal fluid from optical coherence tomography (OCT) scans is crucial for diagnosing and monitoring retinal diseases.
• Existing methods face challenges due to noise, motion artifacts, and variations in fluid pocket characteristics within OCT images.

Purpose of the Study:

• To develop and evaluate a fully convolutional neural network (FCNN) for the automated segmentation of sub-retinal fluid in OCT scans.
• To specifically address the segmentation of sub-retinal fluid in the context of central serous chorioretinopathy (CSC) pathology.

Main Methods:

• A fully convolutional neural network (FCNN) architecture was employed to automatically learn relevant features for segmentation.
• The FCNN was trained and evaluated on a dataset comprising 15 OCT volumes of patients with CSC.
• Performance was quantified using standard metrics including Dice score, Precision, and Recall.

Main Results:

• The proposed FCNN achieved an average Dice score of 0.91, indicating high overlap between segmented and actual fluid regions.
• The method demonstrated strong performance with an average Precision of 0.93 and Recall of 0.89.
• The FCNN effectively handled noise and motion artifacts common in OCT imaging.

Conclusions:

• Fully convolutional neural networks are well-suited for automated retinal fluid segmentation tasks in OCT imaging.
• The developed FCNN provides an accurate and robust method for segmenting sub-retinal fluid in CSC, outperforming previous approaches.
• This automated technique holds potential for improving clinical diagnosis and management of CSC and related retinal conditions.